Amazon AWS Graviton3 benchmarks by Michael Larabel.
c7g.4xlarge Graviton3 Processor: ARMv8 Neoverse-V1 (16 Cores), Motherboard: Amazon EC2 c7g.4xlarge (1.0 BIOS), Chipset: Amazon Device 0200, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (aarch64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -vJava Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected
a1.4xlarge Graviton Changed Processor to ARMv8 Cortex-A72 (16 Cores) .
Changed Motherboard to Amazon EC2 a1.4xlarge (1.0 BIOS) .
Security Change: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Not affected + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of Branch predictor hardening BHB + srbds: Not affected + tsx_async_abort: Not affected
c6g.4xlarge Graviton2 Changed Processor to ARMv8 Neoverse-N1 (16 Cores) .
Changed Motherboard to Amazon EC2 c6g.4xlarge (1.0 BIOS) .
c6a.4xlarge EPYC Processor: AMD EPYC 7R13 (8 Cores / 16 Threads) , Motherboard: Amazon EC2 c6a.4xlarge (1.0 BIOS) , Chipset: Intel 440FX 82441FX PMC , Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (x86_64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -vProcessor Notes: CPU Microcode: 0xa001144Java Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional IBRS_FW STIBP: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected
c6i.4xlarge Xeon Changed Processor to Intel Xeon Platinum 8375C (8 Cores / 16 Threads) .
Changed Motherboard to Amazon EC2 c6i.4xlarge (1.0 BIOS) .
Processor Change: CPU Microcode: 0xd000331Security Change: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected
Amazon EC2 Graviton3 Benchmark Comparison Processor Motherboard Chipset Memory Disk Network OS Kernel Compiler File-System System Layer c7g.4xlarge Graviton3 a1.4xlarge Graviton c6g.4xlarge Graviton2 c6a.4xlarge EPYC c6i.4xlarge Xeon ARMv8 Neoverse-V1 (16 Cores) Amazon EC2 c7g.4xlarge (1.0 BIOS) Amazon Device 0200 32GB 193GB Amazon Elastic Block Store Amazon Elastic Ubuntu 22.04 5.15.0-1004-aws (aarch64) GCC 11.2.0 ext4 amazon ARMv8 Cortex-A72 (16 Cores) Amazon EC2 a1.4xlarge (1.0 BIOS) ARMv8 Neoverse-N1 (16 Cores) Amazon EC2 c6g.4xlarge (1.0 BIOS) AMD EPYC 7R13 (8 Cores / 16 Threads) Amazon EC2 c6a.4xlarge (1.0 BIOS) Intel 440FX 82441FX PMC 5.15.0-1004-aws (x86_64) Intel Xeon Platinum 8375C (8 Cores / 16 Threads) Amazon EC2 c6i.4xlarge (1.0 BIOS) OpenBenchmarking.org Kernel Details - Transparent Huge Pages: madvise Compiler Details - c7g.4xlarge Graviton3: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -v - a1.4xlarge Graviton: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -v - c6g.4xlarge Graviton2: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -v - c6a.4xlarge EPYC: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v - c6i.4xlarge Xeon: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v Java Details - OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1) Python Details - Python 3.10.4 Security Details - c7g.4xlarge Graviton3: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected - a1.4xlarge Graviton: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Not affected + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of Branch predictor hardening BHB + srbds: Not affected + tsx_async_abort: Not affected - c6g.4xlarge Graviton2: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected - c6a.4xlarge EPYC: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional IBRS_FW STIBP: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected - c6i.4xlarge Xeon: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected Processor Details - c6a.4xlarge EPYC: CPU Microcode: 0xa001144 - c6i.4xlarge Xeon: CPU Microcode: 0xd000331
c7g.4xlarge Graviton3 a1.4xlarge Graviton c6g.4xlarge Graviton2 c6a.4xlarge EPYC c6i.4xlarge Xeon Logarithmic Result Overview Phoronix Test Suite High Performance Conjugate Gradient Algebraic Multi-Grid Benchmark ACES DGEMM ONNX Runtime Xcompact3d Incompact3d NAS Parallel Benchmarks Timed MrBayes Analysis GPAW LULESH GROMACS Apache HTTP Server simdjson TensorFlow Lite ASTC Encoder Timed Node.js Compilation LAMMPS Molecular Dynamics Simulator PyBench PHPBench libavif avifenc Timed ImageMagick Compilation Timed Apache Compilation Timed LLVM Compilation Rodinia OpenSSL Zstd Compression SecureMark Ngspice Liquid-DSP Build2 Timed PHP Compilation DaCapo Benchmark WebP Image Encode Timed Gem5 Compilation nginx C-Ray TSCP Stockfish POV-Ray 7-Zip Compression Stress-NG asmFish Google SynthMark LeelaChessZero Coremark N-Queens m-queens
c7g.4xlarge Graviton3 a1.4xlarge Graviton c6g.4xlarge Graviton2 c6a.4xlarge EPYC c6i.4xlarge Xeon Logarithmic Per Dollar Result Overview Phoronix Test Suite LeelaChessZero PHPBench Stockfish ONNX Runtime Algebraic Multi-Grid Benchmark ACES DGEMM Zstd Compression asmFish NAS Parallel Benchmarks 7-Zip Compression simdjson Apache HTTP Server Coremark OpenSSL Liquid-DSP High Performance Conjugate Gradient LULESH GROMACS SecureMark nginx Stress-NG Google SynthMark P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M P.D.G.M
Amazon EC2 Graviton3 Benchmark Comparison stress-ng: Memory Copying npb: MG.C npb: CG.C npb: SP.C compress-zstd: 3 - Compression Speed npb: FT.C hpcg: amg: incompact3d: input.i3d 129 Cells Per Direction mt-dgemm: Sustained Floating-Point Rate incompact3d: input.i3d 193 Cells Per Direction stress-ng: CPU Stress simdjson: DistinctUserID mrbayes: Primate Phylogeny Analysis npb: IS.D tensorflow-lite: Mobilenet Float gpaw: Carbon Nanotube avifenc: 2 simdjson: PartialTweets lulesh: apache: 100 astcenc: Thorough gromacs: MPI CPU - water_GMX50_bare tensorflow-lite: Inception V4 apache: 500 apache: 200 simdjson: Kostya npb: BT.C openssl: RSA4096 tensorflow-lite: Inception ResNet V2 apache: 1000 tensorflow-lite: SqueezeNet astcenc: Exhaustive rodinia: OpenMP CFD Solver openssl: RSA4096 avifenc: 0 build-nodejs: Time To Compile lammps: Rhodopsin Protein pybench: Total For Average Test Times phpbench: PHP Benchmark Suite build-imagemagick: Time To Compile tensorflow-lite: NASNet Mobile build-apache: Time To Compile dacapobench: Jython build-llvm: Ninja npb: EP.D ngspice: C2670 dacapobench: Tradesoap simdjson: LargeRand securemark: SecureMark-TLS dacapobench: Tradebeans liquid-dsp: 16 - 256 - 57 build2: Time To Compile build-php: Time To Compile compress-7zip: Compression Rating ngspice: C7552 webp: Quality 100, Lossless, Highest Compression build-gem5: Time To Compile webp: Quality 100, Lossless avifenc: 6, Lossless nginx: 1000 nginx: 500 nginx: 200 compress-zstd: 19 - Decompression Speed nginx: 100 tscp: AI Chess Performance compress-zstd: 19, Long Mode - Decompression Speed stockfish: Total Time rodinia: OpenMP LavaMD povray: Trace Time compress-zstd: 19, Long Mode - Compression Speed compress-zstd: 19 - Compression Speed dacapobench: H2 stress-ng: Crypto asmfish: 1024 Hash Memory, 26 Depth synthmark: VoiceMark_100 openssl: SHA256 stress-ng: Vector Math npb: LU.C lczero: Eigen lczero: BLAS coremark: CoreMark Size 666 - Iterations Per Second n-queens: Elapsed Time compress-7zip: Decompression Rating m-queens: Time To Solve stress-ng: IO_uring onnx: super-resolution-10 - CPU - Standard onnx: ArcFace ResNet-100 - CPU - Standard onnx: fcn-resnet101-11 - CPU - Standard onnx: bertsquad-12 - CPU - Standard onnx: GPT-2 - CPU - Standard tensorflow-lite: Mobilenet Quant c-ray: Total Time - 4K, 16 Rays Per Pixel rodinia: OpenMP Streamcluster c7g.4xlarge Graviton3 a1.4xlarge Graviton c6g.4xlarge Graviton2 c6a.4xlarge EPYC c6i.4xlarge Xeon 6693.32 13481.61 6571.95 4467.19 4639.1 11791.77 26.3058 1258807333 8.01671425 5.853864 29.1258570 5029.71 2.69 251.397 1041.90 2156.60 155.180 141.698 2.62 10940.939 67231.88 13.9248 1.128 41855.1 73546.32 73676.95 1.94 10339.53 2546.4 40051.3 72719.33 3257.94 139.3797 10.478 178460.4 256.841 497.579 11.291 1185 666484 27.904 11591.9 26.940 3940 544.929 934.72 198.224 3524 0.7 183708 3203 383606667 115.020 69.483 97824 191.286 48.208 391.171 22.769 11.908 346814.75 346613.34 352380.98 3050.3 345710.87 1370094 3240.6 27608891 143.334 37.863 39.5 41.2 2951 23181.81 32134123 675.635 13722045973 55258.17 7730.41 1189 1103 405413.860554 21.536 73054 66.822 843015.78 2817 609 38 407 7990 1502.95 38.517 13.296 798.24 3266.36 1213.15 1293.80 633.9 2927.16 3.77834 186716933 53.7706274 0.891391 182.583939 2366.00 0.8 644.788 197.57 9990.15 769.346 449.022 0.78 2328.2724 18636.43 33.5198 0.316 188910 20133.49 20887.58 0.63 3148.18 588.3 171169 19278.68 12014.7 277.7669 41.450 45328.6 768.302 1765.910 3.245 3452 241259 93.632 30986.7 74.742 12997 1784.600 339.20 473.901 11182 0.3 74356 9045 165513333 353.912 196.029 32498 480.793 124.708 1155.615 61.801 33.991 138205.11 139414.84 141436.20 1121.7 143155.48 538500 1213.9 10980430 360.304 93.801 16 16.9 6740 11985.38 15331550 331.070 6785689517 27341.47 2558.12 128 135 203869.402017 32.285 40891 110.368 918172.37 757 165 10 115 2312 5724.66 104.761 47.430 2903.00 6720.68 3520.86 2356.16 2878.8 6244.48 19.7218 932652900 11.5733547 4.785123 41.0240835 3404.94 1.53 384.753 372.76 2500.87 215.528 238.205 1.51 6016.1627 46995.35 16.5222 0.781 46793.9 50077.81 50059.97 1.19 6449.11 660.6 45955.7 46629.45 3969.35 159.2039 17.035 53951.5 406.937 628.401 7.935 1741 449855 40.333 14985.4 34.201 5626 682.981 558.88 263.724 4506 0.49 120301 4344 262890000 142.277 88.897 71285 255.205 66.147 488.805 31.082 16.518 308213.13 310596.58 308938.67 2051.6 307349.36 872313 2196.3 21679245 215.666 51.047 31.0 34.6 3964 17924.18 26540482 470.389 10723184083 37753.89 5133.89 834 864 315464.339800 23.136 59445 75.224 770521.81 2072 334 28 322 6948 1980.24 62.323 15.484 3551.80 16826.43 6169.22 8094.79 2768.7 18299.96 5.06042 267670700 28.2797661 2.432432 110.770027 13304.50 4.30 120.636 541.35 2159.72 302.956 93.946 3.64 5452.1051 77567.69 7.9818 1.004 44920.6 81995.64 83070.00 2.80 13134.46 2088.9 41366.6 71537.11 3103.12 72.3908 21.789 136784.2 195.532 664.347 5.067 1961 480741 32.626 9266.86 23.532 4616 760.344 466.21 245.886 4052 0.95 213288 3167 509746667 150.994 67.084 62562 180.356 48.677 515.201 26.708 16.394 388657.76 389030.11 390932.79 2907.5 388010.76 1442631 2826.0 23857623 224.331 49.435 25.9 30.0 3019 13556.06 26187688 663.073 11691403353 53787.61 25140.55 1001 1091 345133.440541 16.378 57318 72.330 768723.46 3696 1192 65 488 5617 3847.96 69.349 18.383 3150.49 26298.81 9522.82 9563.22 3440.6 20423.57 8.66031 661364767 17.8682772 2.230545 69.2169978 12527.16 4.30 134.924 861.57 1965.07 202.106 97.735 3.71 8112.3715 86545.57 7.2625 1.452 41185.7 91746.57 94458.22 2.46 13888.40 2161.3 41179.7 79830.96 2983.93 69.6387 20.446 140964.4 204.994 604.620 6.220 997 828186 29.737 10900.6 22.527 4013 685.704 1103.22 147.893 3815 0.86 230549 2928 373100000 136.801 64.337 66631 161.081 41.805 469.940 21.122 17.529 347345.49 351672.92 356829.93 2582.0 356302.84 1272596 2666.1 22081961 281.389 52.784 33.8 38.1 2921 10210.34 23746200 565.690 7096993937 40140.30 38136.77 1466 1397 285378.841661 18.839 45653 91.231 1037943.37 3450 1374 139 773 7944 3967.39 92.545 23.512 OpenBenchmarking.org
Stress-NG Stress-NG is a Linux stress tool developed by Colin King of Canonical. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Bogo Ops/s, More Is Better Stress-NG 0.14 Test: Memory Copying c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1400 2800 4200 5600 7000 SE +/- 0.94, N = 3 SE +/- 11.57, N = 3 SE +/- 3.75, N = 3 SE +/- 0.91, N = 3 SE +/- 3.52, N = 3 3150.49 3551.80 2903.00 798.24 6693.32 1. (CC) gcc options: -O2 -std=gnu99 -lm -lapparmor -latomic -lc -lcrypt -ldl -ljpeg -lrt -lz -pthread
Performance / Cost
OpenBenchmarking.org Bogo Ops/s Per Dollar, More Is Better Stress-NG 0.14 Performance / Cost - Test: Memory Copying c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300K 600K 900K 1200K 1500K 525081.67 710360.00 580600.00 266080.00 1338664.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: MG.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6K 12K 18K 24K 30K SE +/- 184.24, N = 3 SE +/- 30.62, N = 3 SE +/- 1.39, N = 3 SE +/- 1.64, N = 3 SE +/- 4.69, N = 3 26298.81 16826.43 6720.68 3266.36 13481.61 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: MG.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M 26298810.00 8413215.00 1680170.00 544393.33 6740805.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 7 seconds per run: cost approximately 0.001 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 11 seconds per run: cost approximately 0.002 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 25 seconds per run: cost approximately 0.004 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 52 seconds per run: cost approximately 0.006 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 13 seconds per run: cost approximately 0.002 dollars.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: CG.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 66.44, N = 3 SE +/- 81.25, N = 3 SE +/- 9.95, N = 3 SE +/- 11.79, N = 6 SE +/- 17.12, N = 3 9522.82 6169.22 3520.86 1213.15 6571.95 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: CG.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 700K 1400K 2100K 2800K 3500K 3174273.33 1542305.00 586810.00 86653.57 1642987.50 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 16 seconds per run: cost approximately 0.003 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.004 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 42 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: SP.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 73.65, N = 3 SE +/- 24.63, N = 3 SE +/- 0.57, N = 3 SE +/- 2.51, N = 3 SE +/- 9.61, N = 3 9563.22 8094.79 2356.16 1293.80 4467.19 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: SP.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 70K 140K 210K 280K 350K 329766.21 261122.26 25065.53 10107.81 84286.60 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 153 seconds per run: cost approximately 0.029 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.031 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 10 minutes, 19 seconds per run: cost approximately 0.094 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 18 minutes, 47 seconds per run: cost approximately 0.128 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 5 minutes, 26 seconds per run: cost approximately 0.053 dollars.
Zstd Compression This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MB/s, More Is Better Zstd Compression 1.5.0 Compression Level: 3 - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1000 2000 3000 4000 5000 SE +/- 29.53, N = 3 SE +/- 18.93, N = 3 SE +/- 3.74, N = 3 SE +/- 4.47, N = 3 SE +/- 9.57, N = 3 3440.6 2768.7 2878.8 633.9 4639.1 -llzma -llzma -llzma -llzma 1. (CC) gcc options: -O3 -pthread -lz
Performance / Cost
OpenBenchmarking.org MB/s Per Dollar, More Is Better Zstd Compression 1.5.0 Performance / Cost - Compression Level: 3 - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 170K 340K 510K 680K 850K 491514.29 397714.29 481383.33 126780.00 773183.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 38 seconds per run: cost approximately 0.007 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 39 seconds per run: cost approximately 0.007 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 39 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 42 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 37 seconds per run: cost approximately 0.006 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: FT.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4K 8K 12K 16K 20K SE +/- 40.24, N = 3 SE +/- 45.90, N = 3 SE +/- 1.10, N = 3 SE +/- 1.73, N = 3 SE +/- 1.17, N = 3 20423.57 18299.96 6244.48 2927.16 11791.77 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: FT.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1.1M 2.2M 3.3M 4.4M 5.5M 5105892.50 4574990.00 624448.00 182947.50 1965295.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 21 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 67 seconds per run: cost approximately 0.01 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 143 seconds per run: cost approximately 0.016 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 36 seconds per run: cost approximately 0.006 dollars.
High Performance Conjugate Gradient HPCG is the High Performance Conjugate Gradient and is a new scientific benchmark from Sandia National Lans focused for super-computer testing with modern real-world workloads compared to HPCC. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GFLOP/s, More Is Better High Performance Conjugate Gradient 3.1 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6 12 18 24 30 SE +/- 0.04033, N = 3 SE +/- 0.00225, N = 3 SE +/- 0.01639, N = 3 SE +/- 0.00065, N = 3 SE +/- 0.03738, N = 3 8.66031 5.06042 19.72180 3.77834 26.30580 1. (CXX) g++ options: -O3 -ffast-math -ftree-vectorize -lmpi_cxx -lmpi
Performance / Cost
OpenBenchmarking.org GFLOP/s Per Dollar, More Is Better High Performance Conjugate Gradient 3.1 Performance / Cost c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 400 800 1200 1600 2000 360.85 202.42 1160.11 107.95 1753.72 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 125 seconds per run: cost approximately 0.024 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 148 seconds per run: cost approximately 0.025 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 111 seconds per run: cost approximately 0.017 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 5 minutes, 8 seconds per run: cost approximately 0.035 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 96 seconds per run: cost approximately 0.015 dollars.
Algebraic Multi-Grid Benchmark AMG is a parallel algebraic multigrid solver for linear systems arising from problems on unstructured grids. The driver provided with AMG builds linear systems for various 3-dimensional problems. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Figure Of Merit, More Is Better Algebraic Multi-Grid Benchmark 1.2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300M 600M 900M 1200M 1500M SE +/- 5114517.12, N = 3 SE +/- 103921.81, N = 3 SE +/- 3420043.89, N = 3 SE +/- 176548.39, N = 3 SE +/- 952437.28, N = 3 661364767 267670700 932652900 186716933 1258807333 1. (CC) gcc options: -lparcsr_ls -lparcsr_mv -lseq_mv -lIJ_mv -lkrylov -lHYPRE_utilities -lm -fopenmp -lmpi
Performance / Cost
OpenBenchmarking.org Figure Of Merit Per Dollar, More Is Better Algebraic Multi-Grid Benchmark 1.2 Performance / Cost c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300000M 600000M 900000M 1200000M 1500000M 330682383500.00 89223566666.67 466326450000.00 37343386600.00 1258807333000.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 8 seconds per run: cost approximately 0.002 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 17 seconds per run: cost approximately 0.003 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 12 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 48 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 8 seconds per run: cost approximately 0.001 dollars.
Xcompact3d Incompact3d Xcompact3d Incompact3d is a Fortran-MPI based, finite difference high-performance code for solving the incompressible Navier-Stokes equation and as many as you need scalar transport equations. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Xcompact3d Incompact3d 2021-03-11 Input: input.i3d 129 Cells Per Direction c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 12 24 36 48 60 SE +/- 0.09619197, N = 3 SE +/- 0.03718674, N = 3 SE +/- 0.01351889, N = 3 SE +/- 0.02862870, N = 3 SE +/- 0.01401446, N = 3 17.86827720 28.27976610 11.57335470 53.77062740 8.01671425 1. (F9X) gfortran options: -cpp -O2 -funroll-loops -floop-optimize -fcray-pointer -fbacktrace -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Xcompact3d Incompact3d 2021-03-11 Performance / Cost - Input: input.i3d 129 Cells Per Direction c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0727 0.1454 0.2181 0.2908 0.3635 0.054 0.141 0.023 0.323 0.008 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 18 seconds per run: cost approximately 0.003 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 29 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 12 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 55 seconds per run: cost approximately 0.006 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 8 seconds per run: cost approximately 0.001 dollars.
ACES DGEMM This is a multi-threaded DGEMM benchmark. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GFLOP/s, More Is Better ACES DGEMM 1.0 Sustained Floating-Point Rate c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1.3171 2.6342 3.9513 5.2684 6.5855 SE +/- 0.003819, N = 3 SE +/- 0.023324, N = 6 SE +/- 0.007139, N = 3 SE +/- 0.002370, N = 3 SE +/- 0.016350, N = 3 2.230545 2.432432 4.785123 0.891391 5.853864 1. (CC) gcc options: -O3 -march=native -fopenmp
Performance / Cost
OpenBenchmarking.org GFLOP/s Per Dollar, More Is Better ACES DGEMM 1.0 Performance / Cost - Sustained Floating-Point Rate c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200 400 600 800 1000 111.53 152.03 683.59 30.74 975.64 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 104 seconds per run: cost approximately 0.02 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 95 seconds per run: cost approximately 0.016 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 49 seconds per run: cost approximately 0.007 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 4 minutes, 20 seconds per run: cost approximately 0.029 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 40 seconds per run: cost approximately 0.006 dollars.
Xcompact3d Incompact3d Xcompact3d Incompact3d is a Fortran-MPI based, finite difference high-performance code for solving the incompressible Navier-Stokes equation and as many as you need scalar transport equations. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Xcompact3d Incompact3d 2021-03-11 Input: input.i3d 193 Cells Per Direction c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40 80 120 160 200 SE +/- 0.14, N = 3 SE +/- 0.12, N = 3 SE +/- 0.01, N = 3 SE +/- 0.15, N = 3 SE +/- 0.03, N = 3 69.22 110.77 41.02 182.58 29.13 1. (F9X) gfortran options: -cpp -O2 -funroll-loops -floop-optimize -fcray-pointer -fbacktrace -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Xcompact3d Incompact3d 2021-03-11 Performance / Cost - Input: input.i3d 193 Cells Per Direction c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.8627 1.7254 2.5881 3.4508 4.3135 0.900 2.105 0.246 3.834 0.146 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 70 seconds per run: cost approximately 0.013 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 112 seconds per run: cost approximately 0.019 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 42 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 184 seconds per run: cost approximately 0.021 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
Stress-NG Stress-NG is a Linux stress tool developed by Colin King of Canonical. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Bogo Ops/s, More Is Better Stress-NG 0.14 Test: CPU Stress c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3K 6K 9K 12K 15K SE +/- 155.66, N = 3 SE +/- 37.60, N = 3 SE +/- 0.54, N = 3 SE +/- 0.16, N = 3 SE +/- 0.41, N = 3 12527.16 13304.50 3404.94 2366.00 5029.71 1. (CC) gcc options: -O2 -std=gnu99 -lm -lapparmor -latomic -lc -lcrypt -ldl -ljpeg -lrt -lz -pthread
Performance / Cost
OpenBenchmarking.org Bogo Ops/s Per Dollar, More Is Better Stress-NG 0.14 Performance / Cost - Test: CPU Stress c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 600K 1200K 1800K 2400K 3000K 2087860.00 2660900.00 680988.00 788666.67 1005942.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
simdjson This is a benchmark of SIMDJSON, a high performance JSON parser. SIMDJSON aims to be the fastest JSON parser and is used by projects like Microsoft FishStore, Yandex ClickHouse, Shopify, and others. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GB/s, More Is Better simdjson 1.0 Throughput Test: DistinctUserID c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.9675 1.935 2.9025 3.87 4.8375 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 4.30 4.30 1.53 0.80 2.69 1. (CXX) g++ options: -O3
Performance / Cost
OpenBenchmarking.org GB/s Per Dollar, More Is Better simdjson 1.0 Performance / Cost - Throughput Test: DistinctUserID c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 70 140 210 280 350 330.77 286.67 191.25 133.33 224.17 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 67 seconds per run: cost approximately 0.013 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 88 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 50 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 55 seconds per run: cost approximately 0.006 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 73 seconds per run: cost approximately 0.012 dollars.
Timed MrBayes Analysis This test performs a bayesian analysis of a set of primate genome sequences in order to estimate their phylogeny. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed MrBayes Analysis 3.2.7 Primate Phylogeny Analysis c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 140 280 420 560 700 SE +/- 1.43, N = 3 SE +/- 0.35, N = 3 SE +/- 0.11, N = 3 SE +/- 0.49, N = 3 SE +/- 0.24, N = 3 134.92 120.64 384.75 644.79 251.40 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msha -maes -mavx -mfma -mavx2 -mavx512f -mavx512cd -mavx512vl -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -mrdrnd -mbmi -mbmi2 -madx -mabm -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4a -msha -maes -mavx -mfma -mavx2 -mrdrnd -mbmi -mbmi2 -madx -mabm 1. (CC) gcc options: -O3 -std=c99 -pedantic -lm
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed MrBayes Analysis 3.2.7 Performance / Cost - Primate Phylogeny Analysis c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 11 22 33 44 55 3.508 2.533 22.316 47.070 10.056 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 135 seconds per run: cost approximately 0.026 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.021 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 6 minutes, 25 seconds per run: cost approximately 0.058 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 10 minutes, 45 seconds per run: cost approximately 0.073 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 4 minutes, 11 seconds per run: cost approximately 0.04 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: IS.D c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200 400 600 800 1000 SE +/- 2.14, N = 3 SE +/- 0.47, N = 3 SE +/- 0.20, N = 3 SE +/- 0.31, N = 3 SE +/- 2.29, N = 3 861.57 541.35 372.76 197.57 1041.90 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: IS.D c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K 78324.55 41642.31 23297.50 8980.46 115766.67 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 59 seconds per run: cost approximately 0.011 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 76 seconds per run: cost approximately 0.013 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 103 seconds per run: cost approximately 0.016 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 195 seconds per run: cost approximately 0.022 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.009 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: Mobilenet Float c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 1.81, N = 3 SE +/- 1.03, N = 3 SE +/- 28.63, N = 3 SE +/- 113.94, N = 3 SE +/- 19.61, N = 3 1965.07 2159.72 2500.87 9990.15 2156.60
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: Mobilenet Float c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 16 32 48 64 80 23.58 21.60 22.51 69.93 21.57 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
GPAW GPAW is a density-functional theory (DFT) Python code based on the projector-augmented wave (PAW) method and the atomic simulation environment (ASE). Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better GPAW 22.1 Input: Carbon Nanotube c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 170 340 510 680 850 SE +/- 0.24, N = 3 SE +/- 0.17, N = 3 SE +/- 0.13, N = 3 SE +/- 5.37, N = 3 SE +/- 0.08, N = 3 202.11 302.96 215.53 769.35 155.18 1. (CC) gcc options: -shared -fwrapv -O2 -lxc -lblas -lmpi
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better GPAW 22.1 Performance / Cost - Input: Carbon Nanotube c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 15 30 45 60 75 7.680 15.754 7.112 66.933 3.880 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 203 seconds per run: cost approximately 0.038 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 5 minutes, 4 seconds per run: cost approximately 0.052 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 216 seconds per run: cost approximately 0.033 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 12 minutes, 51 seconds per run: cost approximately 0.087 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 156 seconds per run: cost approximately 0.025 dollars.
libavif avifenc This is a test of the AOMedia libavif library testing the encoding of a JPEG image to AV1 Image Format (AVIF). Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better libavif avifenc 0.10 Encoder Speed: 2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 100 200 300 400 500 SE +/- 0.26, N = 3 SE +/- 0.44, N = 3 SE +/- 0.12, N = 3 SE +/- 0.29, N = 3 SE +/- 0.11, N = 3 97.74 93.95 238.21 449.02 141.70 1. (CXX) g++ options: -O3 -fPIC -lm
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better libavif avifenc 0.10 Performance / Cost - Encoder Speed: 2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 5 10 15 20 25 1.857 1.503 8.575 22.900 3.259 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 98 seconds per run: cost approximately 0.019 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 94 seconds per run: cost approximately 0.016 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 238 seconds per run: cost approximately 0.036 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 7 minutes, 29 seconds per run: cost approximately 0.051 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 142 seconds per run: cost approximately 0.023 dollars.
simdjson This is a benchmark of SIMDJSON, a high performance JSON parser. SIMDJSON aims to be the fastest JSON parser and is used by projects like Microsoft FishStore, Yandex ClickHouse, Shopify, and others. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GB/s, More Is Better simdjson 1.0 Throughput Test: PartialTweets c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.8348 1.6696 2.5044 3.3392 4.174 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 3.71 3.64 1.51 0.78 2.62 1. (CXX) g++ options: -O3
Performance / Cost
OpenBenchmarking.org GB/s Per Dollar, More Is Better simdjson 1.0 Performance / Cost - Throughput Test: PartialTweets c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 60 120 180 240 300 285.39 242.67 188.75 130.00 218.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 69 seconds per run: cost approximately 0.013 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 86 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 50 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.006 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 73 seconds per run: cost approximately 0.012 dollars.
LULESH LULESH is the Livermore Unstructured Lagrangian Explicit Shock Hydrodynamics. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org z/s, More Is Better LULESH 2.0.3 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 14.20, N = 3 SE +/- 5.52, N = 3 SE +/- 4.88, N = 3 SE +/- 6.27, N = 3 SE +/- 76.73, N = 3 8112.37 5452.11 6016.16 2328.27 10940.94 1. (CXX) g++ options: -O3 -fopenmp -lm -lmpi_cxx -lmpi
Performance / Cost
OpenBenchmarking.org z/s Per Dollar, More Is Better LULESH 2.0.3 Performance / Cost c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2M 4M 6M 8M 10M 8112371.50 2726052.55 6016162.70 776090.80 10940939.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 7 seconds per run: cost approximately 0.001 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 9 seconds per run: cost approximately 0.001 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 5 seconds per run: cost approximately 0.001 dollars.
Apache HTTP Server This is a test of the Apache HTTPD web server. This Apache HTTPD web server benchmark test profile makes use of the Golang "Bombardier" program for facilitating the HTTP requests over a fixed period time with a configurable number of concurrent clients. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Requests Per Second, More Is Better Apache HTTP Server 2.4.48 Concurrent Requests: 100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K SE +/- 389.13, N = 3 SE +/- 211.56, N = 3 SE +/- 93.03, N = 3 SE +/- 28.97, N = 3 SE +/- 38.09, N = 3 86545.57 77567.69 46995.35 18636.43 67231.88 1. (CC) gcc options: -shared -fPIC -O2
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better Apache HTTP Server 2.4.48 Performance / Cost - Concurrent Requests: 100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1.1M 2.2M 3.3M 4.4M 5.5M 5090915.88 5171179.33 3356810.71 1863643.00 4482125.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
ASTC Encoder ASTC Encoder (astcenc) is for the Adaptive Scalable Texture Compression (ASTC) format commonly used with OpenGL, OpenGL ES, and Vulkan graphics APIs. This test profile does a coding test of both compression/decompression. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better ASTC Encoder 3.2 Preset: Thorough c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 8 16 24 32 40 SE +/- 0.0001, N = 3 SE +/- 0.0154, N = 3 SE +/- 0.0064, N = 3 SE +/- 0.0061, N = 3 SE +/- 0.0011, N = 3 7.2625 7.9818 16.5222 33.5198 13.9248 1. (CXX) g++ options: -O3 -flto -pthread
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better ASTC Encoder 3.2 Performance / Cost - Preset: Thorough c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0378 0.0756 0.1134 0.1512 0.189 0.015 0.016 0.050 0.168 0.042 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 13 seconds per run: cost approximately 0.002 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 13 seconds per run: cost approximately 0.002 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 44 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 19 seconds per run: cost approximately 0.003 dollars.
GROMACS The GROMACS (GROningen MAchine for Chemical Simulations) molecular dynamics package testing with the water_GMX50 data. This test profile allows selecting between CPU and GPU-based GROMACS builds. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Ns Per Day, More Is Better GROMACS 2022.1 Implementation: MPI CPU - Input: water_GMX50_bare c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.3267 0.6534 0.9801 1.3068 1.6335 SE +/- 0.001, N = 3 SE +/- 0.002, N = 3 SE +/- 0.001, N = 3 SE +/- 0.000, N = 3 SE +/- 0.002, N = 3 1.452 1.004 0.781 0.316 1.128 1. (CXX) g++ options: -O3
Performance / Cost
OpenBenchmarking.org Ns Per Day Per Dollar, More Is Better GROMACS 2022.1 Performance / Cost - Implementation: MPI CPU - Input: water_GMX50_bare c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 14 28 42 56 70 60.500 33.467 22.971 5.016 45.120 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 125 seconds per run: cost approximately 0.024 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 178 seconds per run: cost approximately 0.03 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 226 seconds per run: cost approximately 0.034 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 9 minutes, 18 seconds per run: cost approximately 0.063 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 157 seconds per run: cost approximately 0.025 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: Inception V4 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40K 80K 120K 160K 200K SE +/- 75.14, N = 3 SE +/- 53.95, N = 3 SE +/- 197.89, N = 3 SE +/- 1746.17, N = 3 SE +/- 210.27, N = 3 41185.7 44920.6 46793.9 188910.0 41855.1
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: Inception V4 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300 600 900 1200 1500 494.23 449.21 421.15 1322.37 418.55 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 62 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
Apache HTTP Server This is a test of the Apache HTTPD web server. This Apache HTTPD web server benchmark test profile makes use of the Golang "Bombardier" program for facilitating the HTTP requests over a fixed period time with a configurable number of concurrent clients. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Requests Per Second, More Is Better Apache HTTP Server 2.4.48 Concurrent Requests: 500 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K SE +/- 833.50, N = 7 SE +/- 636.46, N = 13 SE +/- 578.32, N = 3 SE +/- 93.64, N = 3 SE +/- 89.82, N = 3 91746.57 81995.64 50077.81 20133.49 73546.32 1. (CC) gcc options: -shared -fPIC -O2
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better Apache HTTP Server 2.4.48 Performance / Cost - Concurrent Requests: 500 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1.2M 2.4M 3.6M 4.8M 6M 5396857.06 5466376.00 3576986.43 2013349.00 4903088.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
Result
OpenBenchmarking.org Requests Per Second, More Is Better Apache HTTP Server 2.4.48 Concurrent Requests: 200 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K SE +/- 615.05, N = 3 SE +/- 644.29, N = 3 SE +/- 112.65, N = 3 SE +/- 59.55, N = 3 SE +/- 649.31, N = 3 94458.22 83070.00 50059.97 20887.58 73676.95 1. (CC) gcc options: -shared -fPIC -O2
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better Apache HTTP Server 2.4.48 Performance / Cost - Concurrent Requests: 200 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1.2M 2.4M 3.6M 4.8M 6M 5556365.88 5538000.00 3575712.14 2088758.00 4911796.67 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
simdjson This is a benchmark of SIMDJSON, a high performance JSON parser. SIMDJSON aims to be the fastest JSON parser and is used by projects like Microsoft FishStore, Yandex ClickHouse, Shopify, and others. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GB/s, More Is Better simdjson 1.0 Throughput Test: Kostya c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.63 1.26 1.89 2.52 3.15 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 2.46 2.80 1.19 0.63 1.94 1. (CXX) g++ options: -O3
Performance / Cost
OpenBenchmarking.org GB/s Per Dollar, More Is Better simdjson 1.0 Performance / Cost - Throughput Test: Kostya c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 60 120 180 240 300 223.64 280.00 119.00 63.00 215.56 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.011 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 58 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 63 seconds per run: cost approximately 0.01 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 84 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.009 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: BT.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3K 6K 9K 12K 15K SE +/- 22.04, N = 3 SE +/- 98.45, N = 3 SE +/- 3.20, N = 3 SE +/- 3.44, N = 3 SE +/- 7.36, N = 3 13888.40 13134.46 6449.11 3148.18 10339.53 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: BT.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80K 160K 240K 320K 400K 356112.82 354985.41 94839.85 30270.96 229767.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 208 seconds per run: cost approximately 0.039 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 220 seconds per run: cost approximately 0.037 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 7 minutes, 28 seconds per run: cost approximately 0.068 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 15 minutes, 18 seconds per run: cost approximately 0.104 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 4 minutes, 40 seconds per run: cost approximately 0.045 dollars.
OpenSSL OpenSSL is an open-source toolkit that implements SSL (Secure Sockets Layer) and TLS (Transport Layer Security) protocols. This test profile makes use of the built-in "openssl speed" benchmarking capabilities. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org sign/s, More Is Better OpenSSL 3.0 Algorithm: RSA4096 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 500 1000 1500 2000 2500 SE +/- 4.47, N = 3 SE +/- 1.40, N = 3 SE +/- 0.03, N = 3 SE +/- 0.12, N = 3 SE +/- 0.23, N = 3 2161.3 2088.9 660.6 588.3 2546.4 -m64 -m64 1. (CC) gcc options: -pthread -O3 -lssl -lcrypto -ldl
Performance / Cost
OpenBenchmarking.org sign/s Per Dollar, More Is Better OpenSSL 3.0 Performance / Cost - Algorithm: RSA4096 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 50K 100K 150K 200K 250K 196481.82 208890.00 73400.00 84042.86 254640.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.011 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.01 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: Inception ResNet V2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40K 80K 120K 160K 200K SE +/- 110.01, N = 3 SE +/- 27.66, N = 3 SE +/- 336.95, N = 3 SE +/- 825.35, N = 3 SE +/- 305.31, N = 3 41179.7 41366.6 45955.7 171169.0 40051.3
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: Inception ResNet V2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300 600 900 1200 1500 494.16 413.67 413.60 1198.18 400.51 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
Apache HTTP Server This is a test of the Apache HTTPD web server. This Apache HTTPD web server benchmark test profile makes use of the Golang "Bombardier" program for facilitating the HTTP requests over a fixed period time with a configurable number of concurrent clients. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Requests Per Second, More Is Better Apache HTTP Server 2.4.48 Concurrent Requests: 1000 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K SE +/- 335.63, N = 3 SE +/- 397.88, N = 3 SE +/- 276.10, N = 3 SE +/- 98.61, N = 3 SE +/- 83.83, N = 3 79830.96 71537.11 46629.45 19278.68 72719.33 1. (CC) gcc options: -shared -fPIC -O2
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better Apache HTTP Server 2.4.48 Performance / Cost - Concurrent Requests: 1000 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1000K 2000K 3000K 4000K 5000K 4695938.82 4769140.67 3330675.00 1927868.00 4847955.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 91 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 91 seconds per run: cost approximately 0.015 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: SqueezeNet c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3K 6K 9K 12K 15K SE +/- 3.54, N = 3 SE +/- 1.37, N = 3 SE +/- 37.23, N = 3 SE +/- 46.48, N = 3 SE +/- 22.07, N = 3 2983.93 3103.12 3969.35 12014.70 3257.94
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: SqueezeNet c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 35.81 31.03 35.72 84.10 32.58 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
ASTC Encoder ASTC Encoder (astcenc) is for the Adaptive Scalable Texture Compression (ASTC) format commonly used with OpenGL, OpenGL ES, and Vulkan graphics APIs. This test profile does a coding test of both compression/decompression. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better ASTC Encoder 3.2 Preset: Exhaustive c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 60 120 180 240 300 SE +/- 0.04, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 SE +/- 0.07, N = 3 SE +/- 0.01, N = 3 69.64 72.39 159.20 277.77 139.38 1. (CXX) g++ options: -O3 -flto -pthread
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better ASTC Encoder 3.2 Performance / Cost - Preset: Exhaustive c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3 6 9 12 15 0.975 0.941 3.980 9.166 3.206 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 75 seconds per run: cost approximately 0.014 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 77 seconds per run: cost approximately 0.013 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 166 seconds per run: cost approximately 0.025 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 4 minutes, 48 seconds per run: cost approximately 0.033 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 144 seconds per run: cost approximately 0.023 dollars.
Rodinia Rodinia is a suite focused upon accelerating compute-intensive applications with accelerators. CUDA, OpenMP, and OpenCL parallel models are supported by the included applications. This profile utilizes select OpenCL, NVIDIA CUDA and OpenMP test binaries at the moment. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenMP CFD Solver c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 9 18 27 36 45 SE +/- 0.02, N = 3 SE +/- 0.08, N = 3 SE +/- 0.05, N = 3 SE +/- 0.08, N = 3 SE +/- 0.02, N = 3 20.45 21.79 17.04 41.45 10.48 1. (CXX) g++ options: -O2 -lOpenCL
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Rodinia 3.1 Performance / Cost - Test: OpenMP CFD Solver c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0466 0.0932 0.1398 0.1864 0.233 0.082 0.087 0.051 0.207 0.021 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 22 seconds per run: cost approximately 0.004 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 17 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 41 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars.
OpenSSL OpenSSL is an open-source toolkit that implements SSL (Secure Sockets Layer) and TLS (Transport Layer Security) protocols. This test profile makes use of the built-in "openssl speed" benchmarking capabilities. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org verify/s, More Is Better OpenSSL 3.0 Algorithm: RSA4096 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40K 80K 120K 160K 200K SE +/- 47.94, N = 3 SE +/- 74.60, N = 3 SE +/- 3.30, N = 3 SE +/- 63.75, N = 3 SE +/- 82.61, N = 3 140964.4 136784.2 53951.5 45328.6 178460.4 -m64 -m64 1. (CC) gcc options: -pthread -O3 -lssl -lcrypto -ldl
Performance / Cost
OpenBenchmarking.org verify/s Per Dollar, More Is Better OpenSSL 3.0 Performance / Cost - Algorithm: RSA4096 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4M 8M 12M 16M 20M 12814945.46 13678420.00 5994611.11 6475514.29 17846040.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.011 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 60 seconds per run: cost approximately 0.01 dollars.
libavif avifenc This is a test of the AOMedia libavif library testing the encoding of a JPEG image to AV1 Image Format (AVIF). Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better libavif avifenc 0.10 Encoder Speed: 0 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 170 340 510 680 850 SE +/- 0.33, N = 3 SE +/- 0.62, N = 3 SE +/- 0.13, N = 3 SE +/- 0.58, N = 3 SE +/- 0.18, N = 3 204.99 195.53 406.94 768.30 256.84 1. (CXX) g++ options: -O3 -fPIC -lm
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better libavif avifenc 0.10 Performance / Cost - Encoder Speed: 0 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 15 30 45 60 75 7.995 6.453 25.230 66.842 10.530 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 205 seconds per run: cost approximately 0.039 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 196 seconds per run: cost approximately 0.033 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 6 minutes, 47 seconds per run: cost approximately 0.062 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 12 minutes, 48 seconds per run: cost approximately 0.087 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 4 minutes, 17 seconds per run: cost approximately 0.041 dollars.
Timed Node.js Compilation This test profile times how long it takes to build/compile Node.js itself from source. Node.js is a JavaScript run-time built from the Chrome V8 JavaScript engine while itself is written in C/C++. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed Node.js Compilation 17.3 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 400 800 1200 1600 2000 SE +/- 0.42, N = 3 SE +/- 0.26, N = 3 SE +/- 0.37, N = 3 SE +/- 1.80, N = 3 SE +/- 2.06, N = 3 604.62 664.35 628.40 1765.91 497.58
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed Node.js Compilation 17.3 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80 160 240 320 400 68.93 75.07 59.70 353.18 39.81 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 10 minutes, 5 seconds per run: cost approximately 0.114 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 11 minutes, 4 seconds per run: cost approximately 0.113 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 10 minutes, 28 seconds per run: cost approximately 0.095 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 29 minutes, 26 seconds per run: cost approximately 0.2 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 8 minutes, 18 seconds per run: cost approximately 0.08 dollars.
LAMMPS Molecular Dynamics Simulator LAMMPS is a classical molecular dynamics code, and an acronym for Large-scale Atomic/Molecular Massively Parallel Simulator. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org ns/day, More Is Better LAMMPS Molecular Dynamics Simulator 29Oct2020 Model: Rhodopsin Protein c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3 6 9 12 15 SE +/- 0.009, N = 3 SE +/- 0.039, N = 12 SE +/- 0.014, N = 3 SE +/- 0.040, N = 3 SE +/- 0.060, N = 3 6.220 5.067 7.935 3.245 11.291 1. (CXX) g++ options: -O3 -lm
Performance / Cost
OpenBenchmarking.org ns/day Per Dollar, More Is Better LAMMPS Molecular Dynamics Simulator 29Oct2020 Performance / Cost - Model: Rhodopsin Protein c6i.4xlarge Xeon c6a.4xlarge EPYC a1.4xlarge Graviton 1300 2600 3900 5200 6500 6220.00 5067.00 3245.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 3 seconds per run: cost approximately 0.001 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 4 seconds per run: cost approximately 0.001 dollars. 3. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 6 seconds per run: cost approximately 0.001 dollars.
PyBench This test profile reports the total time of the different average timed test results from PyBench. PyBench reports average test times for different functions such as BuiltinFunctionCalls and NestedForLoops, with this total result providing a rough estimate as to Python's average performance on a given system. This test profile runs PyBench each time for 20 rounds. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Milliseconds, Fewer Is Better PyBench 2018-02-16 Total For Average Test Times c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 700 1400 2100 2800 3500 SE +/- 3.84, N = 3 SE +/- 1.53, N = 3 SE +/- 1.67, N = 3 SE +/- 18.15, N = 3 SE +/- 0.33, N = 3 997 1961 1741 3452 1185
Performance / Cost
OpenBenchmarking.org Milliseconds x Dollar, Fewer Is Better PyBench 2018-02-16 Performance / Cost - Total For Average Test Times c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 7 14 21 28 35 3.988 15.688 10.446 31.068 4.740 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 47 seconds per run: cost approximately 0.008 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 40 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 79 seconds per run: cost approximately 0.009 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 27 seconds per run: cost approximately 0.004 dollars.
PHPBench PHPBench is a benchmark suite for PHP. It performs a large number of simple tests in order to bench various aspects of the PHP interpreter. PHPBench can be used to compare hardware, operating systems, PHP versions, PHP accelerators and caches, compiler options, etc. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Score, More Is Better PHPBench 0.8.1 PHP Benchmark Suite c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200K 400K 600K 800K 1000K SE +/- 983.65, N = 3 SE +/- 2681.41, N = 3 SE +/- 743.13, N = 3 SE +/- 816.27, N = 3 SE +/- 525.83, N = 3 828186 480741 449855 241259 666484
Performance / Cost
OpenBenchmarking.org Score Per Dollar, More Is Better PHPBench 0.8.1 Performance / Cost - PHP Benchmark Suite c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40M 80M 120M 160M 200M 165637200.00 68677285.71 64265000.00 26806555.56 133296800.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.005 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 42 seconds per run: cost approximately 0.007 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 44 seconds per run: cost approximately 0.007 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 83 seconds per run: cost approximately 0.009 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
Timed ImageMagick Compilation This test times how long it takes to build ImageMagick. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed ImageMagick Compilation 6.9.0 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 SE +/- 0.07, N = 3 SE +/- 0.06, N = 3 SE +/- 0.22, N = 3 SE +/- 0.27, N = 3 SE +/- 0.13, N = 3 29.74 32.63 40.33 93.63 27.90
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed ImageMagick Compilation 6.9.0 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.2318 0.4636 0.6954 0.9272 1.159 0.178 0.196 0.242 1.030 0.140 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 33 seconds per run: cost approximately 0.006 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 40 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 94 seconds per run: cost approximately 0.011 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 28 seconds per run: cost approximately 0.005 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: NASNet Mobile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 7K 14K 21K 28K 35K SE +/- 166.62, N = 14 SE +/- 23.44, N = 3 SE +/- 203.15, N = 15 SE +/- 49.84, N = 3 SE +/- 121.56, N = 15 10900.60 9266.86 14985.40 30986.70 11591.90
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: NASNet Mobile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 50 100 150 200 250 130.81 92.67 134.87 216.91 115.92 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
Timed Apache Compilation This test times how long it takes to build the Apache HTTPD web server. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed Apache Compilation 2.4.41 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 SE +/- 0.05, N = 3 SE +/- 0.07, N = 3 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 SE +/- 0.05, N = 3 22.53 23.53 34.20 74.74 26.94
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed Apache Compilation 2.4.41 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.1514 0.3028 0.4542 0.6056 0.757 0.090 0.094 0.171 0.673 0.108 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.004 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 34 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 75 seconds per run: cost approximately 0.009 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 27 seconds per run: cost approximately 0.004 dollars.
DaCapo Benchmark This test runs the DaCapo Benchmarks written in Java and intended to test system/CPU performance. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org msec, Fewer Is Better DaCapo Benchmark 9.12-MR1 Java Test: Jython c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3K 6K 9K 12K 15K SE +/- 24.07, N = 4 SE +/- 23.33, N = 4 SE +/- 23.29, N = 4 SE +/- 48.38, N = 4 SE +/- 6.99, N = 4 4013 4616 5626 12997 3940
Performance / Cost
OpenBenchmarking.org msec x Dollar, Fewer Is Better DaCapo Benchmark 9.12-MR1 Performance / Cost - Java Test: Jython c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6 12 18 24 30 4.013 4.616 5.626 25.994 3.940 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 6 seconds per run: cost approximately 0.001 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 6 seconds per run: cost approximately 0.001 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 8 seconds per run: cost approximately 0.001 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 18 seconds per run: cost approximately 0.002 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 6 seconds per run: cost approximately 0.001 dollars.
Timed LLVM Compilation This test times how long it takes to build the LLVM compiler stack. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed LLVM Compilation 13.0 Build System: Ninja c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 400 800 1200 1600 2000 SE +/- 0.12, N = 3 SE +/- 0.05, N = 3 SE +/- 0.49, N = 3 SE +/- 0.34, N = 3 SE +/- 5.19, N = 3 685.70 760.34 682.98 1784.60 544.93
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed LLVM Compilation 13.0 Performance / Cost - Build System: Ninja c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80 160 240 320 400 89.14 98.08 70.35 360.49 47.95 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 11 minutes, 26 seconds per run: cost approximately 0.13 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 12 minutes, 40 seconds per run: cost approximately 0.129 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 11 minutes, 23 seconds per run: cost approximately 0.103 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 29 minutes, 45 seconds per run: cost approximately 0.202 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 9 minutes, 5 seconds per run: cost approximately 0.088 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: EP.D c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200 400 600 800 1000 SE +/- 19.93, N = 9 SE +/- 0.06, N = 3 SE +/- 0.23, N = 3 SE +/- 0.24, N = 3 SE +/- 0.39, N = 3 1103.22 466.21 558.88 339.20 934.72 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: EP.D c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 10K 20K 30K 40K 50K 45967.50 9324.20 15104.87 7373.91 38946.67 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 125 seconds per run: cost approximately 0.024 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 4 minutes, 55 seconds per run: cost approximately 0.05 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 4 minutes, 6 seconds per run: cost approximately 0.037 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 6 minutes, 46 seconds per run: cost approximately 0.046 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 147 seconds per run: cost approximately 0.024 dollars.
Ngspice Ngspice is an open-source SPICE circuit simulator. Ngspice was originally based on the Berkeley SPICE electronic circuit simulator. Ngspice supports basic threading using OpenMP. This test profile is making use of the ISCAS 85 benchmark circuits. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Ngspice 34 Circuit: C2670 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 100 200 300 400 500 SE +/- 1.80, N = 4 SE +/- 1.17, N = 3 SE +/- 0.91, N = 3 SE +/- 3.48, N = 3 SE +/- 0.86, N = 3 147.89 245.89 263.72 473.90 198.22 1. (CC) gcc options: -O0 -fopenmp -lm -lstdc++ -lfftw3 -lXaw -lXmu -lXt -lXext -lX11 -lXft -lfontconfig -lXrender -lfreetype -lSM -lICE
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Ngspice 34 Performance / Cost - Circuit: C2670 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6 12 18 24 30 4.141 10.327 10.549 25.591 6.343 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 149 seconds per run: cost approximately 0.028 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 4 minutes, 7 seconds per run: cost approximately 0.042 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 4 minutes, 25 seconds per run: cost approximately 0.04 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 7 minutes, 57 seconds per run: cost approximately 0.054 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 199 seconds per run: cost approximately 0.032 dollars.
DaCapo Benchmark This test runs the DaCapo Benchmarks written in Java and intended to test system/CPU performance. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org msec, Fewer Is Better DaCapo Benchmark 9.12-MR1 Java Test: Tradesoap c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 24.39, N = 4 SE +/- 16.15, N = 4 SE +/- 27.95, N = 4 SE +/- 71.92, N = 4 SE +/- 14.95, N = 4 3815 4052 4506 11182 3524
Performance / Cost
OpenBenchmarking.org msec x Dollar, Fewer Is Better DaCapo Benchmark 9.12-MR1 Performance / Cost - Java Test: Tradesoap c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 8 16 24 32 40 7.630 8.104 9.012 33.546 7.048 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 13 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 26 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars.
simdjson This is a benchmark of SIMDJSON, a high performance JSON parser. SIMDJSON aims to be the fastest JSON parser and is used by projects like Microsoft FishStore, Yandex ClickHouse, Shopify, and others. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org GB/s, More Is Better simdjson 1.0 Throughput Test: LargeRandom c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.2138 0.4276 0.6414 0.8552 1.069 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 0.86 0.95 0.49 0.30 0.70 1. (CXX) g++ options: -O3
Performance / Cost
OpenBenchmarking.org GB/s Per Dollar, More Is Better simdjson 1.0 Performance / Cost - Throughput Test: LargeRandom c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 86.00 105.56 54.44 37.50 77.78 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 54 seconds per run: cost approximately 0.01 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 52 seconds per run: cost approximately 0.009 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 57 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 69 seconds per run: cost approximately 0.008 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 53 seconds per run: cost approximately 0.009 dollars.
SecureMark SecureMark is an objective, standardized benchmarking framework for measuring the efficiency of cryptographic processing solutions developed by EEMBC. SecureMark-TLS is benchmarking Transport Layer Security performance with a focus on IoT/edge computing. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org marks, More Is Better SecureMark 1.0.4 Benchmark: SecureMark-TLS c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 50K 100K 150K 200K 250K SE +/- 864.34, N = 3 SE +/- 3310.19, N = 9 SE +/- 23.07, N = 3 SE +/- 59.40, N = 3 SE +/- 773.26, N = 3 230549 213288 120301 74356 183708 1. (CC) gcc options: -pedantic -O3
Performance / Cost
OpenBenchmarking.org marks Per Dollar, More Is Better SecureMark 1.0.4 Performance / Cost - Benchmark: SecureMark-TLS c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 900K 1800K 2700K 3600K 4500K 4349981.13 4352816.33 2864309.52 2323625.00 4082400.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 4 minutes, 42 seconds per run: cost approximately 0.053 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 4 minutes, 50 seconds per run: cost approximately 0.049 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 4 minutes, 40 seconds per run: cost approximately 0.042 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 4 minutes, 41 seconds per run: cost approximately 0.032 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 4 minutes, 41 seconds per run: cost approximately 0.045 dollars.
DaCapo Benchmark This test runs the DaCapo Benchmarks written in Java and intended to test system/CPU performance. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org msec, Fewer Is Better DaCapo Benchmark 9.12-MR1 Java Test: Tradebeans c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 19.24, N = 20 SE +/- 23.53, N = 11 SE +/- 40.13, N = 4 SE +/- 44.35, N = 4 SE +/- 26.73, N = 4 2928 3167 4344 9045 3203
Performance / Cost
OpenBenchmarking.org msec x Dollar, Fewer Is Better DaCapo Benchmark 9.12-MR1 Performance / Cost - Java Test: Tradebeans c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6 12 18 24 30 5.856 6.334 8.688 27.135 3.203 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 9 seconds per run: cost approximately 0.002 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 9 seconds per run: cost approximately 0.002 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 12 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 9 seconds per run: cost approximately 0.001 dollars.
Liquid-DSP LiquidSDR's Liquid-DSP is a software-defined radio (SDR) digital signal processing library. This test profile runs a multi-threaded benchmark of this SDR/DSP library focused on embedded platform usage. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org samples/s, More Is Better Liquid-DSP 2021.01.31 Threads: 16 - Buffer Length: 256 - Filter Length: 57 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 110M 220M 330M 440M 550M SE +/- 41633.32, N = 3 SE +/- 489364.67, N = 3 SE +/- 35118.85, N = 3 SE +/- 8819.17, N = 3 SE +/- 400097.21, N = 3 373100000 509746667 262890000 165513333 383606667 1. (CC) gcc options: -O3 -pthread -lm -lc -lliquid
Performance / Cost
OpenBenchmarking.org samples/s Per Dollar, More Is Better Liquid-DSP 2021.01.31 Performance / Cost - Threads: 16 - Buffer Length: 256 - Filter Length: 57 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40000M 80000M 120000M 160000M 200000M 93275000000.00 169915555666.67 87630000000.00 82756666500.00 127868889000.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.003 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.002 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.003 dollars.
Build2 This test profile measures the time to bootstrap/install the build2 C++ build toolchain from source. Build2 is a cross-platform build toolchain for C/C++ code and features Cargo-like features. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Build2 0.13 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80 160 240 320 400 SE +/- 0.69, N = 3 SE +/- 0.87, N = 3 SE +/- 0.70, N = 3 SE +/- 1.89, N = 3 SE +/- 0.64, N = 3 136.80 150.99 142.28 353.91 115.02
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Build2 0.13 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4 8 12 16 20 3.557 3.926 2.988 14.156 2.185 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 137 seconds per run: cost approximately 0.026 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 151 seconds per run: cost approximately 0.026 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 142 seconds per run: cost approximately 0.021 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 5 minutes, 54 seconds per run: cost approximately 0.04 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 115 seconds per run: cost approximately 0.019 dollars.
Timed PHP Compilation This test times how long it takes to build PHP 7. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed PHP Compilation 7.4.2 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40 80 120 160 200 SE +/- 0.09, N = 3 SE +/- 0.05, N = 3 SE +/- 0.31, N = 3 SE +/- 0.08, N = 3 SE +/- 0.11, N = 3 64.34 67.08 88.90 196.03 69.48
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed PHP Compilation 7.4.2 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.9704 1.9408 2.9112 3.8816 4.852 0.772 0.738 1.156 4.313 0.764 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 64 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 67 seconds per run: cost approximately 0.011 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 89 seconds per run: cost approximately 0.013 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 196 seconds per run: cost approximately 0.022 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 69 seconds per run: cost approximately 0.011 dollars.
7-Zip Compression This is a test of 7-Zip compression/decompression with its integrated benchmark feature. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MIPS, More Is Better 7-Zip Compression 21.06 Test: Compression Rating c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20K 40K 60K 80K 100K SE +/- 174.34, N = 3 SE +/- 16.02, N = 3 SE +/- 44.77, N = 3 SE +/- 91.00, N = 3 SE +/- 159.36, N = 3 66631 62562 71285 32498 97824 1. (CXX) g++ options: -lpthread -ldl -O2 -fPIC
Performance / Cost
OpenBenchmarking.org MIPS Per Dollar, More Is Better 7-Zip Compression 21.06 Performance / Cost - Test: Compression Rating c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 5M 10M 15M 20M 25M 11105166.67 12512400.00 17821250.00 6499600.00 24456000.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 34 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 29 seconds per run: cost approximately 0.004 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 48 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars.
Ngspice Ngspice is an open-source SPICE circuit simulator. Ngspice was originally based on the Berkeley SPICE electronic circuit simulator. Ngspice supports basic threading using OpenMP. This test profile is making use of the ISCAS 85 benchmark circuits. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Ngspice 34 Circuit: C7552 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 100 200 300 400 500 SE +/- 0.33, N = 3 SE +/- 0.66, N = 3 SE +/- 2.40, N = 7 SE +/- 1.19, N = 3 SE +/- 1.94, N = 3 161.08 180.36 255.21 480.79 191.29 1. (CC) gcc options: -O0 -fopenmp -lm -lstdc++ -lfftw3 -lXaw -lXmu -lXt -lXext -lX11 -lXft -lfontconfig -lXrender -lfreetype -lSM -lICE
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Ngspice 34 Performance / Cost - Circuit: C7552 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6 12 18 24 30 4.994 5.771 10.208 26.924 6.121 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 165 seconds per run: cost approximately 0.031 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 186 seconds per run: cost approximately 0.032 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 4 minutes, 23 seconds per run: cost approximately 0.04 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 8 minutes, 13 seconds per run: cost approximately 0.056 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 196 seconds per run: cost approximately 0.032 dollars.
WebP Image Encode This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Encode Time - Seconds, Fewer Is Better WebP Image Encode 1.1 Encode Settings: Quality 100, Lossless, Highest Compression c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 30 60 90 120 150 SE +/- 0.34, N = 3 SE +/- 0.06, N = 3 SE +/- 0.01, N = 3 SE +/- 0.09, N = 3 SE +/- 0.01, N = 3 41.81 48.68 66.15 124.71 48.21 -ltiff -ltiff -ltiff -ltiff 1. (CC) gcc options: -fvisibility=hidden -O2 -lm -ljpeg -lpng16
Performance / Cost
OpenBenchmarking.org Encode Time - Seconds x Dollar, Fewer Is Better WebP Image Encode 1.1 Performance / Cost - Encode Settings: Quality 100, Lossless, Highest Compression c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.3929 0.7858 1.1787 1.5716 1.9645 0.334 0.389 0.661 1.746 0.386 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 42 seconds per run: cost approximately 0.008 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 49 seconds per run: cost approximately 0.008 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 66 seconds per run: cost approximately 0.01 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 125 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 48 seconds per run: cost approximately 0.008 dollars.
Timed Gem5 Compilation This test times how long it takes to compile Gem5. Gem5 is a simulator for computer system architecture research. Gem5 is widely used for computer architecture research within the industry, academia, and more. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Timed Gem5 Compilation 21.2 Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200 400 600 800 1000 SE +/- 0.59, N = 3 SE +/- 0.79, N = 3 SE +/- 0.53, N = 3 SE +/- 0.78, N = 3 SE +/- 1.33, N = 3 469.94 515.20 488.81 1155.62 391.17
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Timed Gem5 Compilation 21.2 Performance / Cost - Time To Compile c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 30 60 90 120 150 41.83 45.34 36.17 151.39 24.64 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 7 minutes, 50 seconds per run: cost approximately 0.089 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 8 minutes, 35 seconds per run: cost approximately 0.088 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 8 minutes, 9 seconds per run: cost approximately 0.074 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 19 minutes, 16 seconds per run: cost approximately 0.131 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 6 minutes, 31 seconds per run: cost approximately 0.063 dollars.
WebP Image Encode This is a test of Google's libwebp with the cwebp image encode utility and using a sample 6000x4000 pixel JPEG image as the input. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Encode Time - Seconds, Fewer Is Better WebP Image Encode 1.1 Encode Settings: Quality 100, Lossless c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 14 28 42 56 70 SE +/- 0.03, N = 3 SE +/- 0.17, N = 15 SE +/- 0.02, N = 3 SE +/- 0.06, N = 3 SE +/- 0.09, N = 3 21.12 26.71 31.08 61.80 22.77 -ltiff -ltiff -ltiff -ltiff 1. (CC) gcc options: -fvisibility=hidden -O2 -lm -ljpeg -lpng16
Performance / Cost
OpenBenchmarking.org Encode Time - Seconds x Dollar, Fewer Is Better WebP Image Encode 1.1 Performance / Cost - Encode Settings: Quality 100, Lossless c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0974 0.1948 0.2922 0.3896 0.487 0.084 0.134 0.155 0.433 0.091 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 21 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 27 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 31 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 62 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars.
libavif avifenc This is a test of the AOMedia libavif library testing the encoding of a JPEG image to AV1 Image Format (AVIF). Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better libavif avifenc 0.10 Encoder Speed: 6, Lossless c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 8 16 24 32 40 SE +/- 0.03, N = 3 SE +/- 0.12, N = 3 SE +/- 0.17, N = 3 SE +/- 0.31, N = 3 SE +/- 0.01, N = 3 17.53 16.39 16.52 33.99 11.91 1. (CXX) g++ options: -O3 -fPIC -lm
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better libavif avifenc 0.10 Performance / Cost - Encoder Speed: 6, Lossless c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0306 0.0612 0.0918 0.1224 0.153 0.053 0.049 0.050 0.136 0.024 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 18 seconds per run: cost approximately 0.003 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 16 seconds per run: cost approximately 0.003 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 17 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 34 seconds per run: cost approximately 0.004 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 12 seconds per run: cost approximately 0.002 dollars.
nginx This is a benchmark of the lightweight Nginx HTTP(S) web-server. This Nginx web server benchmark test profile makes use of the Golang "Bombardier" program for facilitating the HTTP requests over a fixed period time with a configurable number of concurrent clients. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Requests Per Second, More Is Better nginx 1.21.1 Concurrent Requests: 1000 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80K 160K 240K 320K 400K SE +/- 2637.25, N = 3 SE +/- 781.49, N = 3 SE +/- 1677.89, N = 3 SE +/- 66.96, N = 3 SE +/- 1410.11, N = 3 347345.49 388657.76 308213.13 138205.11 346814.75 1. (CC) gcc options: -lcrypt -lz -O3 -march=native
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better nginx 1.21.1 Performance / Cost - Concurrent Requests: 1000 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M 20432087.65 25910517.33 22015223.57 13820511.00 23120983.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
Result
OpenBenchmarking.org Requests Per Second, More Is Better nginx 1.21.1 Concurrent Requests: 500 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80K 160K 240K 320K 400K SE +/- 1620.39, N = 3 SE +/- 771.95, N = 3 SE +/- 3783.68, N = 3 SE +/- 141.15, N = 3 SE +/- 1017.52, N = 3 351672.92 389030.11 310596.58 139414.84 346613.34 1. (CC) gcc options: -lcrypt -lz -O3 -march=native
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better nginx 1.21.1 Performance / Cost - Concurrent Requests: 500 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M 20686642.35 25935340.67 22185470.00 13941484.00 23107556.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
Result
OpenBenchmarking.org Requests Per Second, More Is Better nginx 1.21.1 Concurrent Requests: 200 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80K 160K 240K 320K 400K SE +/- 1582.66, N = 3 SE +/- 1242.81, N = 3 SE +/- 1347.28, N = 3 SE +/- 133.96, N = 3 SE +/- 3986.77, N = 3 356829.93 390932.79 308938.67 141436.20 352380.98 1. (CC) gcc options: -lcrypt -lz -O3 -march=native
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better nginx 1.21.1 Performance / Cost - Concurrent Requests: 200 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M 20989995.88 26062186.00 22067047.86 14143620.00 23492065.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
Zstd Compression This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MB/s, More Is Better Zstd Compression 1.5.0 Compression Level: 19 - Decompression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 700 1400 2100 2800 3500 SE +/- 24.18, N = 3 SE +/- 3.25, N = 3 SE +/- 12.10, N = 3 SE +/- 4.74, N = 3 SE +/- 7.75, N = 3 2582.0 2907.5 2051.6 1121.7 3050.3 -llzma -llzma -llzma -llzma 1. (CC) gcc options: -O3 -pthread -lz
Performance / Cost
OpenBenchmarking.org MB/s Per Dollar, More Is Better Zstd Compression 1.5.0 Performance / Cost - Compression Level: 19 - Decompression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 90K 180K 270K 360K 450K 286888.89 290750.00 256450.00 112170.00 435757.14 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 49 seconds per run: cost approximately 0.009 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 52 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 88 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 46 seconds per run: cost approximately 0.007 dollars.
nginx This is a benchmark of the lightweight Nginx HTTP(S) web-server. This Nginx web server benchmark test profile makes use of the Golang "Bombardier" program for facilitating the HTTP requests over a fixed period time with a configurable number of concurrent clients. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Requests Per Second, More Is Better nginx 1.21.1 Concurrent Requests: 100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80K 160K 240K 320K 400K SE +/- 1727.81, N = 3 SE +/- 436.72, N = 3 SE +/- 3992.58, N = 3 SE +/- 22.67, N = 3 SE +/- 2009.97, N = 3 356302.84 388010.76 307349.36 143155.48 345710.87 1. (CC) gcc options: -lcrypt -lz -O3 -march=native
Performance / Cost
OpenBenchmarking.org Requests Per Second Per Dollar, More Is Better nginx 1.21.1 Performance / Cost - Concurrent Requests: 100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M 20958990.59 25867384.00 21953525.71 14315548.00 23047391.33 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.014 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 90 seconds per run: cost approximately 0.015 dollars.
TSCP This is a performance test of TSCP, Tom Kerrigan's Simple Chess Program, which has a built-in performance benchmark. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Nodes Per Second, More Is Better TSCP 1.81 AI Chess Performance c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300K 600K 900K 1200K 1500K SE +/- 1099.67, N = 5 SE +/- 4180.17, N = 5 SE +/- 338.27, N = 5 SE +/- 196.86, N = 5 SE +/- 0.00, N = 5 1272596 1442631 872313 538500 1370094 1. (CC) gcc options: -O3 -march=native
Zstd Compression This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MB/s, More Is Better Zstd Compression 1.5.0 Compression Level: 19, Long Mode - Decompression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 700 1400 2100 2800 3500 SE +/- 7.82, N = 3 SE +/- 6.53, N = 3 SE +/- 2.93, N = 3 SE +/- 15.28, N = 3 SE +/- 6.93, N = 3 2666.1 2826.0 2196.3 1213.9 3240.6 -llzma -llzma -llzma -llzma 1. (CC) gcc options: -O3 -pthread -lz
Performance / Cost
OpenBenchmarking.org MB/s Per Dollar, More Is Better Zstd Compression 1.5.0 Performance / Cost - Compression Level: 19, Long Mode - Decompression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 90K 180K 270K 360K 450K 266610.00 256909.09 274537.50 121390.00 405075.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 53 seconds per run: cost approximately 0.01 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 63 seconds per run: cost approximately 0.011 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 91 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 47 seconds per run: cost approximately 0.008 dollars.
Stockfish This is a test of Stockfish, an advanced open-source C++11 chess benchmark that can scale up to 512 CPU threads. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Nodes Per Second, More Is Better Stockfish 13 Total Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 6M 12M 18M 24M 30M SE +/- 242448.39, N = 3 SE +/- 149731.77, N = 3 SE +/- 292329.99, N = 3 SE +/- 123749.22, N = 3 SE +/- 153578.64, N = 3 22081961 23857623 21679245 10980430 27608891 -m64 -msse -msse3 -mpopcnt -mavx2 -mavx512f -mavx512bw -mavx512vnni -mavx512dq -mavx512vl -msse4.1 -mssse3 -msse2 -mbmi2 -m64 -msse -msse3 -mpopcnt -mavx2 -msse4.1 -mssse3 -msse2 -mbmi2 1. (CXX) g++ options: -lgcov -lpthread -fno-exceptions -std=c++17 -fprofile-use -fno-peel-loops -fno-tracer -pedantic -O3 -flto -flto=jobserver
Performance / Cost
OpenBenchmarking.org Nodes Per Second Per Dollar, More Is Better Stockfish 13 Performance / Cost - Total Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1200M 2400M 3600M 4800M 6000M 2760245125.00 3976270500.00 3613207500.00 1220047777.78 5521778200.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 40 seconds per run: cost approximately 0.008 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 36 seconds per run: cost approximately 0.006 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 41 seconds per run: cost approximately 0.006 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 79 seconds per run: cost approximately 0.009 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 34 seconds per run: cost approximately 0.005 dollars.
Rodinia Rodinia is a suite focused upon accelerating compute-intensive applications with accelerators. CUDA, OpenMP, and OpenCL parallel models are supported by the included applications. This profile utilizes select OpenCL, NVIDIA CUDA and OpenMP test binaries at the moment. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenMP LavaMD c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 80 160 240 320 400 SE +/- 0.14, N = 3 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 SE +/- 0.07, N = 3 SE +/- 0.15, N = 3 281.39 224.33 215.67 360.30 143.33 1. (CXX) g++ options: -O2 -lOpenCL
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Rodinia 3.1 Performance / Cost - Test: OpenMP LavaMD c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4 8 12 16 20 14.914 8.525 7.117 14.772 3.297 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 4 minutes, 41 seconds per run: cost approximately 0.053 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 224 seconds per run: cost approximately 0.038 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 216 seconds per run: cost approximately 0.033 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 6 minutes per run: cost approximately 0.041 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 143 seconds per run: cost approximately 0.023 dollars.
POV-Ray This is a test of POV-Ray, the Persistence of Vision Raytracer. POV-Ray is used to create 3D graphics using ray-tracing. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better POV-Ray 3.7.0.7 Trace Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 SE +/- 0.12, N = 3 SE +/- 0.18, N = 3 SE +/- 0.00, N = 3 SE +/- 0.94, N = 15 SE +/- 0.01, N = 3 52.78 49.44 51.05 93.80 37.86 -march=native -march=native 1. (CXX) g++ options: -pipe -O3 -ffast-math -R/usr/lib -lXpm -lSM -lICE -lX11 -ltiff -ljpeg -lpng -lz -lrt -lm -lboost_thread -lboost_system
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better POV-Ray 3.7.0.7 Performance / Cost - Trace Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.2322 0.4644 0.6966 0.9288 1.161 0.528 0.445 0.408 1.032 0.227 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 55 seconds per run: cost approximately 0.01 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 52 seconds per run: cost approximately 0.009 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 54 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 99 seconds per run: cost approximately 0.011 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 40 seconds per run: cost approximately 0.006 dollars.
Zstd Compression This test measures the time needed to compress/decompress a sample file (a FreeBSD disk image - FreeBSD-12.2-RELEASE-amd64-memstick.img) using Zstd compression with options for different compression levels / settings. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MB/s, More Is Better Zstd Compression 1.5.0 Compression Level: 19, Long Mode - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 9 18 27 36 45 SE +/- 0.10, N = 3 SE +/- 0.27, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 SE +/- 0.23, N = 3 33.8 25.9 31.0 16.0 39.5 -llzma -llzma -llzma -llzma 1. (CC) gcc options: -O3 -pthread -lz
Performance / Cost
OpenBenchmarking.org MB/s Per Dollar, More Is Better Zstd Compression 1.5.0 Performance / Cost - Compression Level: 19, Long Mode - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1100 2200 3300 4400 5500 3380.00 2354.55 3875.00 1600.00 4937.50 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 53 seconds per run: cost approximately 0.01 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 63 seconds per run: cost approximately 0.011 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 91 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 47 seconds per run: cost approximately 0.008 dollars.
Result
OpenBenchmarking.org MB/s, More Is Better Zstd Compression 1.5.0 Compression Level: 19 - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 9 18 27 36 45 SE +/- 0.40, N = 3 SE +/- 0.21, N = 3 SE +/- 0.06, N = 3 SE +/- 0.03, N = 3 SE +/- 0.00, N = 3 38.1 30.0 34.6 16.9 41.2 -llzma -llzma -llzma -llzma 1. (CC) gcc options: -O3 -pthread -lz
Performance / Cost
OpenBenchmarking.org MB/s Per Dollar, More Is Better Zstd Compression 1.5.0 Performance / Cost - Compression Level: 19 - Compression Speed c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1300 2600 3900 5200 6500 4233.33 3000.00 4325.00 1690.00 5885.71 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 49 seconds per run: cost approximately 0.009 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 56 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 52 seconds per run: cost approximately 0.008 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 88 seconds per run: cost approximately 0.01 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 46 seconds per run: cost approximately 0.007 dollars.
DaCapo Benchmark This test runs the DaCapo Benchmarks written in Java and intended to test system/CPU performance. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org msec, Fewer Is Better DaCapo Benchmark 9.12-MR1 Java Test: H2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1400 2800 4200 5600 7000 SE +/- 32.93, N = 4 SE +/- 27.42, N = 4 SE +/- 45.89, N = 4 SE +/- 63.66, N = 4 SE +/- 32.57, N = 5 2921 3019 3964 6740 2951
Performance / Cost
OpenBenchmarking.org msec x Dollar, Fewer Is Better DaCapo Benchmark 9.12-MR1 Performance / Cost - Java Test: H2 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3 6 9 12 15 2.921 3.019 7.928 13.480 2.951 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 7 seconds per run: cost approximately 0.001 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 7 seconds per run: cost approximately 0.001 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 10 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 19 seconds per run: cost approximately 0.002 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 7 seconds per run: cost approximately 0.001 dollars.
Stress-NG Stress-NG is a Linux stress tool developed by Colin King of Canonical. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Bogo Ops/s, More Is Better Stress-NG 0.14 Test: Crypto c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 5K 10K 15K 20K 25K SE +/- 5.89, N = 3 SE +/- 3.93, N = 3 SE +/- 92.83, N = 3 SE +/- 6.29, N = 3 SE +/- 32.01, N = 3 10210.34 13556.06 17924.18 11985.38 23181.81 1. (CC) gcc options: -O2 -std=gnu99 -lm -lapparmor -latomic -lc -lcrypt -ldl -ljpeg -lrt -lz -pthread
Performance / Cost
OpenBenchmarking.org Bogo Ops/s Per Dollar, More Is Better Stress-NG 0.14 Performance / Cost - Test: Crypto c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1000K 2000K 3000K 4000K 5000K 1701723.33 2711212.00 3584836.00 3995126.67 4636362.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
asmFish This is a test of asmFish, an advanced chess benchmark written in Assembly. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Nodes/second, More Is Better asmFish 2018-07-23 1024 Hash Memory, 26 Depth c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 7M 14M 21M 28M 35M SE +/- 325631.00, N = 3 SE +/- 303648.79, N = 3 SE +/- 359309.26, N = 3 SE +/- 106812.26, N = 3 SE +/- 104795.40, N = 3 23746200 26187688 26540482 15331550 32134123
Performance / Cost
OpenBenchmarking.org Nodes/second Per Dollar, More Is Better asmFish 2018-07-23 Performance / Cost - 1024 Hash Memory, 26 Depth c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300M 600M 900M 1200M 1500M 698417647.06 818365250.00 1061619280.00 438044285.71 1338921791.67 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 178 seconds per run: cost approximately 0.034 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 187 seconds per run: cost approximately 0.032 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 164 seconds per run: cost approximately 0.025 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 5 minutes, 7 seconds per run: cost approximately 0.035 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 149 seconds per run: cost approximately 0.024 dollars.
Google SynthMark SynthMark is a cross platform tool for benchmarking CPU performance under a variety of real-time audio workloads. It uses a polyphonic synthesizer model to provide standardized tests for latency, jitter and computational throughput. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Voices, More Is Better Google SynthMark 20201109 Test: VoiceMark_100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 150 300 450 600 750 SE +/- 2.00, N = 3 SE +/- 7.09, N = 3 SE +/- 0.33, N = 3 SE +/- 0.00, N = 3 SE +/- 0.32, N = 3 565.69 663.07 470.39 331.07 675.64 1. (CXX) g++ options: -lm -lpthread -std=c++11 -Ofast
Performance / Cost
OpenBenchmarking.org Voices Per Dollar, More Is Better Google SynthMark 20201109 Performance / Cost - Test: VoiceMark_100 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 30K 60K 90K 120K 150K 94281.67 132614.60 94077.80 110356.67 135127.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
OpenSSL OpenSSL is an open-source toolkit that implements SSL (Secure Sockets Layer) and TLS (Transport Layer Security) protocols. This test profile makes use of the built-in "openssl speed" benchmarking capabilities. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org byte/s, More Is Better OpenSSL 3.0 Algorithm: SHA256 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 3000M 6000M 9000M 12000M 15000M SE +/- 606684.16, N = 3 SE +/- 8616254.20, N = 3 SE +/- 47755430.47, N = 3 SE +/- 12563225.46, N = 3 SE +/- 7739237.92, N = 3 7096993937 11691403353 10723184083 6785689517 13722045973 -m64 -m64 1. (CC) gcc options: -pthread -O3 -lssl -lcrypto -ldl
Performance / Cost
OpenBenchmarking.org byte/s Per Dollar, More Is Better OpenSSL 3.0 Performance / Cost - Algorithm: SHA256 c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 100000M 200000M 300000M 400000M 500000M 208735115794.12 377142043645.16 397154966037.04 339284475850.00 473173999068.97 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.034 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.031 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.027 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.02 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 180 seconds per run: cost approximately 0.029 dollars.
Stress-NG Stress-NG is a Linux stress tool developed by Colin King of Canonical. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Bogo Ops/s, More Is Better Stress-NG 0.14 Test: Vector Math c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 12K 24K 36K 48K 60K SE +/- 28.50, N = 3 SE +/- 2.46, N = 3 SE +/- 15.72, N = 3 SE +/- 0.49, N = 3 SE +/- 17.05, N = 3 40140.30 53787.61 37753.89 27341.47 55258.17 1. (CC) gcc options: -O2 -std=gnu99 -lm -lapparmor -latomic -lc -lcrypt -ldl -ljpeg -lrt -lz -pthread
Performance / Cost
OpenBenchmarking.org Bogo Ops/s Per Dollar, More Is Better Stress-NG 0.14 Performance / Cost - Test: Vector Math c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2M 4M 6M 8M 10M 6690050.00 10757522.00 7550778.00 9113823.33 11051634.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
NAS Parallel Benchmarks NPB, NAS Parallel Benchmarks, is a benchmark developed by NASA for high-end computer systems. This test profile currently uses the MPI version of NPB. This test profile offers selecting the different NPB tests/problems and varying problem sizes. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Total Mop/s, More Is Better NAS Parallel Benchmarks 3.4 Test / Class: LU.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 8K 16K 24K 32K 40K SE +/- 160.86, N = 3 SE +/- 18.06, N = 3 SE +/- 0.90, N = 3 SE +/- 0.15, N = 3 SE +/- 1.96, N = 3 38136.77 25140.55 5133.89 2558.12 7730.41 1. (F9X) gfortran options: -O3 -march=native -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz 2. Open MPI 4.1.2
Performance / Cost
OpenBenchmarking.org Total Mop/s Per Dollar, More Is Better NAS Parallel Benchmarks 3.4 Performance / Cost - Test / Class: LU.C c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 800K 1600K 2400K 3200K 4000K 3813677.00 1795753.57 84162.13 28111.21 179776.98 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 54 seconds per run: cost approximately 0.01 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 82 seconds per run: cost approximately 0.014 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 6 minutes, 41 seconds per run: cost approximately 0.061 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 13 minutes, 24 seconds per run: cost approximately 0.091 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 4 minutes, 26 seconds per run: cost approximately 0.043 dollars.
LeelaChessZero LeelaChessZero (lc0 / lczero) is a chess engine automated vian neural networks. This test profile can be used for OpenCL, CUDA + cuDNN, and BLAS (CPU-based) benchmarking. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.28 Backend: Eigen c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300 600 900 1200 1500 SE +/- 13.37, N = 3 SE +/- 11.74, N = 9 SE +/- 12.00, N = 3 SE +/- 0.67, N = 3 SE +/- 9.70, N = 3 1466 1001 834 128 1189 1. (CXX) g++ options: -flto -pthread
Performance / Cost
OpenBenchmarking.org Nodes Per Second Per Dollar, More Is Better LeelaChessZero 0.28 Performance / Cost - Backend: Eigen c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 5K 10K 15K 20K 25K 21558.82 16145.16 14892.86 2976.74 20500.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 5 minutes, 59 seconds per run: cost approximately 0.068 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 6 minutes, 2 seconds per run: cost approximately 0.062 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 6 minutes, 9 seconds per run: cost approximately 0.056 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 6 minutes, 18 seconds per run: cost approximately 0.043 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 6 minutes, 1 second per run: cost approximately 0.058 dollars.
Result
OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.28 Backend: BLAS c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300 600 900 1200 1500 SE +/- 12.41, N = 9 SE +/- 12.82, N = 9 SE +/- 10.22, N = 4 SE +/- 0.88, N = 3 SE +/- 6.44, N = 3 1397 1091 864 135 1103 1. (CXX) g++ options: -flto -pthread
Performance / Cost
OpenBenchmarking.org Nodes Per Second Per Dollar, More Is Better LeelaChessZero 0.28 Performance / Cost - Backend: BLAS c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4K 8K 12K 16K 20K 20544.12 17596.77 15709.09 3214.29 19017.24 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 5 minutes, 58 seconds per run: cost approximately 0.068 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 6 minutes, 2 seconds per run: cost approximately 0.062 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 6 minutes, 7 seconds per run: cost approximately 0.055 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 6 minutes, 14 seconds per run: cost approximately 0.042 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 6 minutes, 1 second per run: cost approximately 0.058 dollars.
Coremark This is a test of EEMBC CoreMark processor benchmark. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Iterations/Sec, More Is Better Coremark 1.0 CoreMark Size 666 - Iterations Per Second c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 90K 180K 270K 360K 450K SE +/- 80.93, N = 3 SE +/- 2163.00, N = 3 SE +/- 49.84, N = 3 SE +/- 116.54, N = 3 SE +/- 3211.91, N = 3 285378.84 345133.44 315464.34 203869.40 405413.86 1. (CC) gcc options: -O2 -lrt" -lrt
Performance / Cost
OpenBenchmarking.org Iterations/Sec Per Dollar, More Is Better Coremark 1.0 Performance / Cost - CoreMark Size 666 - Iterations Per Second c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 30M 60M 90M 120M 150M 57075768.33 86283360.14 105154779.93 67956467.34 135137953.52 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 26 seconds per run: cost approximately 0.005 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 22 seconds per run: cost approximately 0.004 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 21 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 20 seconds per run: cost approximately 0.003 dollars.
N-Queens This is a test of the OpenMP version of a test that solves the N-queens problem. The board problem size is 18. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better N-Queens 1.0 Elapsed Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 7 14 21 28 35 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 18.84 16.38 23.14 32.29 21.54 1. (CC) gcc options: -static -fopenmp -O3 -march=native
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better N-Queens 1.0 Performance / Cost - Elapsed Time c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.029 0.058 0.087 0.116 0.145 0.075 0.049 0.069 0.129 0.086 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 19 seconds per run: cost approximately 0.004 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 16 seconds per run: cost approximately 0.003 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.003 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 32 seconds per run: cost approximately 0.004 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 22 seconds per run: cost approximately 0.004 dollars.
7-Zip Compression This is a test of 7-Zip compression/decompression with its integrated benchmark feature. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org MIPS, More Is Better 7-Zip Compression 21.06 Test: Decompression Rating c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 16K 32K 48K 64K 80K SE +/- 35.00, N = 3 SE +/- 142.56, N = 3 SE +/- 239.68, N = 3 SE +/- 31.21, N = 3 SE +/- 12.88, N = 3 45653 57318 59445 40891 73054 1. (CXX) g++ options: -lpthread -ldl -O2 -fPIC
Performance / Cost
OpenBenchmarking.org MIPS Per Dollar, More Is Better 7-Zip Compression 21.06 Performance / Cost - Test: Decompression Rating c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 4M 8M 12M 16M 20M 7608833.33 11463600.00 14861250.00 8178200.00 18263500.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 34 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 29 seconds per run: cost approximately 0.004 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 48 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 23 seconds per run: cost approximately 0.004 dollars.
m-queens A solver for the N-queens problem with multi-threading support via the OpenMP library. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better m-queens 1.2 Time To Solve c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 SE +/- 0.06, N = 3 SE +/- 0.02, N = 3 SE +/- 0.00, N = 3 SE +/- 0.01, N = 3 SE +/- 0.00, N = 3 91.23 72.33 75.22 110.37 66.82 1. (CXX) g++ options: -fopenmp -O2 -march=native
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better m-queens 1.2 Performance / Cost - Time To Solve c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.349 0.698 1.047 1.396 1.745 1.551 0.868 0.827 1.324 0.735 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 91 seconds per run: cost approximately 0.017 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 72 seconds per run: cost approximately 0.012 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 75 seconds per run: cost approximately 0.011 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 110 seconds per run: cost approximately 0.012 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 67 seconds per run: cost approximately 0.011 dollars.
Stress-NG Stress-NG is a Linux stress tool developed by Colin King of Canonical. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Bogo Ops/s, More Is Better Stress-NG 0.14 Test: IO_uring c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 200K 400K 600K 800K 1000K SE +/- 405.56, N = 3 SE +/- 713.03, N = 3 SE +/- 2395.13, N = 3 SE +/- 3840.04, N = 3 SE +/- 614.16, N = 3 1037943.37 768723.46 770521.81 918172.37 843015.78 1. (CC) gcc options: -O2 -std=gnu99 -lm -lapparmor -latomic -lc -lcrypt -ldl -ljpeg -lrt -lz -pthread
Performance / Cost
OpenBenchmarking.org Bogo Ops/s Per Dollar, More Is Better Stress-NG 0.14 Performance / Cost - Test: IO_uring c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 70M 140M 210M 280M 350M 172990561.67 153744692.00 154104362.00 306057456.67 168603156.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.006 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.003 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 30 seconds per run: cost approximately 0.005 dollars.
ONNX Runtime ONNX Runtime is developed by Microsoft and partners as a open-source, cross-platform, high performance machine learning inferencing and training accelerator. This test profile runs the ONNX Runtime with various models available from the ONNX Zoo. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Inferences Per Minute, More Is Better ONNX Runtime 1.11 Model: super-resolution-10 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 800 1600 2400 3200 4000 SE +/- 1.61, N = 3 SE +/- 234.97, N = 12 SE +/- 1.74, N = 3 SE +/- 0.50, N = 3 SE +/- 1.86, N = 3 3450 3696 2072 757 2817 1. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt
Performance / Cost
OpenBenchmarking.org Inferences Per Minute Per Dollar, More Is Better ONNX Runtime 1.11 Performance / Cost - Model: super-resolution-10 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40K 80K 120K 160K 200K 150000.00 184800.00 115111.11 54071.43 148263.16 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 120 seconds per run: cost approximately 0.023 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 120 seconds per run: cost approximately 0.02 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 120 seconds per run: cost approximately 0.018 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 120 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 120 seconds per run: cost approximately 0.019 dollars.
Result
OpenBenchmarking.org Inferences Per Minute, More Is Better ONNX Runtime 1.11 Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 300 600 900 1200 1500 SE +/- 91.51, N = 12 SE +/- 82.60, N = 12 SE +/- 0.17, N = 3 SE +/- 0.50, N = 3 SE +/- 0.00, N = 3 1374 1192 334 165 609 1. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt
Performance / Cost
OpenBenchmarking.org Inferences Per Minute Per Dollar, More Is Better ONNX Runtime 1.11 Performance / Cost - Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 13K 26K 39K 52K 65K 59739.13 56761.91 18555.56 11785.71 32052.63 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.023 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.021 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.018 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 123 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.019 dollars.
Result
OpenBenchmarking.org Inferences Per Minute, More Is Better ONNX Runtime 1.11 Model: fcn-resnet101-11 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 30 60 90 120 150 SE +/- 0.60, N = 3 SE +/- 5.55, N = 12 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 139 65 28 10 38 1. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt
Performance / Cost
OpenBenchmarking.org Inferences Per Minute Per Dollar, More Is Better ONNX Runtime 1.11 Performance / Cost - Model: fcn-resnet101-11 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1300 2600 3900 5200 6500 6043.48 3095.24 1473.68 666.67 1900.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.023 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 123 seconds per run: cost approximately 0.021 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 123 seconds per run: cost approximately 0.019 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 132 seconds per run: cost approximately 0.015 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.02 dollars.
Result
OpenBenchmarking.org Inferences Per Minute, More Is Better ONNX Runtime 1.11 Model: bertsquad-12 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 170 340 510 680 850 SE +/- 50.92, N = 12 SE +/- 0.58, N = 3 SE +/- 0.17, N = 3 SE +/- 0.88, N = 3 SE +/- 0.17, N = 3 773 488 322 115 407 1. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt
Performance / Cost
OpenBenchmarking.org Inferences Per Minute Per Dollar, More Is Better ONNX Runtime 1.11 Performance / Cost - Model: bertsquad-12 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 7K 14K 21K 28K 35K 33608.70 23238.10 17888.89 8214.29 21421.05 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.023 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.021 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.018 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 123 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 121 seconds per run: cost approximately 0.019 dollars.
Result
OpenBenchmarking.org Inferences Per Minute, More Is Better ONNX Runtime 1.11 Model: GPT-2 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 2K 4K 6K 8K 10K SE +/- 322.41, N = 12 SE +/- 75.29, N = 12 SE +/- 3.50, N = 3 SE +/- 2.20, N = 3 SE +/- 2.40, N = 3 7944 5617 6948 2312 7990 1. (CXX) g++ options: -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -flto -fno-fat-lto-objects -ldl -lrt
Performance / Cost
OpenBenchmarking.org Inferences Per Minute Per Dollar, More Is Better ONNX Runtime 1.11 Performance / Cost - Model: GPT-2 - Device: CPU - Executor: Standard c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 90K 180K 270K 360K 450K 345391.30 267476.19 386000.00 165142.86 399500.00 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.023 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.021 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.018 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 124 seconds per run: cost approximately 0.014 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 122 seconds per run: cost approximately 0.02 dollars.
TensorFlow Lite This is a benchmark of the TensorFlow Lite implementation focused on TensorFlow machine learning for mobile, IoT, edge, and other cases. The current Linux support is limited to running on CPUs. This test profile is measuring the average inference time. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Microseconds, Fewer Is Better TensorFlow Lite 2022-05-18 Model: Mobilenet Quant c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 1200 2400 3600 4800 6000 SE +/- 80.05, N = 12 SE +/- 53.31, N = 15 SE +/- 14.44, N = 3 SE +/- 20.90, N = 3 SE +/- 17.76, N = 3 3967.39 3847.96 1980.24 5724.66 1502.95
Performance / Cost
OpenBenchmarking.org Microseconds x Dollar, Fewer Is Better TensorFlow Lite 2022-05-18 Performance / Cost - Model: Mobilenet Quant c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 11 22 33 44 55 47.61 38.48 17.82 40.07 15.03 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.012 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.009 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.007 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 61 seconds per run: cost approximately 0.01 dollars.
C-Ray This is a test of C-Ray, a simple raytracer designed to test the floating-point CPU performance. This test is multi-threaded (16 threads per core), will shoot 8 rays per pixel for anti-aliasing, and will generate a 1600 x 1200 image. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better C-Ray 1.1 Total Time - 4K, 16 Rays Per Pixel c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 20 40 60 80 100 SE +/- 0.04, N = 3 SE +/- 0.77, N = 5 SE +/- 0.03, N = 3 SE +/- 2.00, N = 15 SE +/- 0.02, N = 3 92.55 69.35 62.32 104.76 38.52 1. (CC) gcc options: -lm -lpthread -O3
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better C-Ray 1.1 Performance / Cost - Total Time - 4K, 16 Rays Per Pixel c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.3749 0.7498 1.1247 1.4996 1.8745 1.666 0.832 0.623 1.257 0.231 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 93 seconds per run: cost approximately 0.018 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 70 seconds per run: cost approximately 0.012 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 63 seconds per run: cost approximately 0.01 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 106 seconds per run: cost approximately 0.012 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 39 seconds per run: cost approximately 0.006 dollars.
Rodinia Rodinia is a suite focused upon accelerating compute-intensive applications with accelerators. CUDA, OpenMP, and OpenCL parallel models are supported by the included applications. This profile utilizes select OpenCL, NVIDIA CUDA and OpenMP test binaries at the moment. Learn more via the OpenBenchmarking.org test page.
Result
OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenMP Streamcluster c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 11 22 33 44 55 SE +/- 0.07, N = 3 SE +/- 0.05, N = 3 SE +/- 0.26, N = 15 SE +/- 0.02, N = 3 SE +/- 0.33, N = 12 23.51 18.38 15.48 47.43 13.30 1. (CXX) g++ options: -O2 -lOpenCL
Performance / Cost
OpenBenchmarking.org Seconds x Dollar, Fewer Is Better Rodinia 3.1 Performance / Cost - Test: OpenMP Streamcluster c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 0.0533 0.1066 0.1599 0.2132 0.2665 0.118 0.055 0.031 0.237 0.027 1. c6i.4xlarge Xeon: $0.68 reported cost per hour, test took average of 24 seconds per run: cost approximately 0.005 dollars. 2. c6a.4xlarge EPYC: $0.612 reported cost per hour, test took average of 18 seconds per run: cost approximately 0.003 dollars. 3. c6g.4xlarge Graviton2: $0.544 reported cost per hour, test took average of 15 seconds per run: cost approximately 0.002 dollars. 4. a1.4xlarge Graviton: $0.408 reported cost per hour, test took average of 47 seconds per run: cost approximately 0.005 dollars. 5. c7g.4xlarge Graviton3: $0.58 reported cost per hour, test took average of 13 seconds per run: cost approximately 0.002 dollars.
Geometric Mean Of All Test Results OpenBenchmarking.org Geometric Mean, More Is Better Geometric Mean Of All Test Results Result Composite - Amazon EC2 Graviton3 Benchmark Comparison c6i.4xlarge Xeon c6a.4xlarge EPYC c6g.4xlarge Graviton2 a1.4xlarge Graviton c7g.4xlarge Graviton3 40 80 120 160 200 168.19 152.48 119.30 53.56 170.78
Hide Result
Number Of First Place Finishes c6i.4xlarge Xeon 35 [37.2%] c6a.4xlarge EPYC 16 [17.0%] c7g.4xlarge Graviton3 43 [45.7%] Number Of First Place Finishes Wins - 94 Tests OpenBenchmarking.org
Hide Result
Number Of Last Place Finishes c6i.4xlarge Xeon 1 [1.1%] c6a.4xlarge EPYC 1 [1.1%] a1.4xlarge Graviton 92 [97.9%] Number Of Last Place Finishes Losses - 94 Tests OpenBenchmarking.org
Hide Result
c7g.4xlarge Graviton3 Processor: ARMv8 Neoverse-V1 (16 Cores), Motherboard: Amazon EC2 c7g.4xlarge (1.0 BIOS), Chipset: Amazon Device 0200, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (aarch64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -vJava Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 24 May 2022 11:30 by user ubuntu.
a1.4xlarge Graviton Processor: ARMv8 Cortex-A72 (16 Cores), Motherboard: Amazon EC2 a1.4xlarge (1.0 BIOS), Chipset: Amazon Device 0200, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (aarch64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -vJava Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Not affected + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of Branch predictor hardening BHB + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 25 May 2022 00:29 by user ubuntu.
c6g.4xlarge Graviton2 Processor: ARMv8 Neoverse-N1 (16 Cores), Motherboard: Amazon EC2 c6g.4xlarge (1.0 BIOS), Chipset: Amazon Device 0200, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (aarch64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=aarch64-linux-gnu --disable-libquadmath --disable-libquadmath-support --disable-werror --enable-bootstrap --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-fix-cortex-a53-843419 --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-nls --enable-objc-gc=auto --enable-plugin --enable-shared --enable-threads=posix --host=aarch64-linux-gnu --program-prefix=aarch64-linux-gnu- --target=aarch64-linux-gnu --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-target-system-zlib=auto -vJava Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + spectre_v1: Mitigation of __user pointer sanitization + spectre_v2: Mitigation of CSV2 BHB + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 25 May 2022 13:01 by user ubuntu.
c6a.4xlarge EPYC Processor: AMD EPYC 7R13 (8 Cores / 16 Threads), Motherboard: Amazon EC2 c6a.4xlarge (1.0 BIOS), Chipset: Intel 440FX 82441FX PMC, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (x86_64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -vProcessor Notes: CPU Microcode: 0xa001144Java Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Retpolines IBPB: conditional IBRS_FW STIBP: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 26 May 2022 00:31 by user ubuntu.
c6i.4xlarge Xeon Processor: Intel Xeon Platinum 8375C (8 Cores / 16 Threads), Motherboard: Amazon EC2 c6i.4xlarge (1.0 BIOS), Chipset: Intel 440FX 82441FX PMC, Memory: 32GB, Disk: 193GB Amazon Elastic Block Store, Network: Amazon Elastic
OS: Ubuntu 22.04, Kernel: 5.15.0-1004-aws (x86_64), Compiler: GCC 11.2.0, File-System: ext4, System Layer: amazon
Kernel Notes: Transparent Huge Pages: madviseCompiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-gBFGDP/gcc-11-11.2.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -vProcessor Notes: CPU Microcode: 0xd000331Java Notes: OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.22.04.1)Python Notes: Python 3.10.4Security Notes: itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl and seccomp + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced IBRS IBPB: conditional RSB filling + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 26 May 2022 00:32 by user ubuntu.