rtx3090 AMD Ryzen 9 7950X 16-Core testing with a ASUS ROG CROSSHAIR X670E HERO (0805 BIOS) and MSI NVIDIA GeForce RTX 4090 24GB on Linuxmint 21.1 via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/2301238-NE-2301232NE06&grt&sor .
rtx3090 Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server Display Driver OpenGL OpenCL Vulkan Compiler File-System Screen Resolution rtx3090-gpu rtx4090 AMD Ryzen 9 7950X 16-Core @ 5.88GHz (16 Cores / 32 Threads) ASUS ROG CROSSHAIR X670E HERO (0805 BIOS) AMD Device 14d8 62GB 2000GB Sabrent Rocket 4.0 Plus + 2000GB CT2000P5PSSD8 + 2000GB Samsung SSD 980 PRO 2TB eVGA NVIDIA GeForce RTX 3090 24GB NVIDIA GA102 HD Audio PHL 328E1 + LG ULTRAWIDE Intel I225-V + Intel Wi-Fi 6 AX210/AX211/AX411 Linuxmint 21.1 5.15.0-58-generic (x86_64) Cinnamon 5.6.7 X Server 1.21.1.3 NVIDIA 525.60.11 4.6.0 OpenCL 3.0 CUDA 12.0.89 1.3.224 GCC 11.3.0 ext4 8960x2160 MSI NVIDIA GeForce RTX 4090 24GB NVIDIA Device 22ba NVIDIA 525.78.01 OpenCL 3.0 CUDA 12.0.133 OpenBenchmarking.org Kernel Details - Transparent Huge Pages: madvise Compiler Details - --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-xKiWfi/gcc-11-11.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-xKiWfi/gcc-11-11.3.0/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 Processor Details - Scaling Governor: amd-pstate performance (Boost: Enabled) - CPU Microcode: 0xa601203 Graphics Details - rtx3090-gpu: BAR1 / Visible vRAM Size: 256 MiB - vBIOS Version: 94.02.42.c0.15 - rtx4090: BAR1 / Visible vRAM Size: 256 MiB - vBIOS Version: 95.02.18.80.74 OpenCL Details - rtx3090-gpu: GPU Compute Cores: 10496 - rtx4090: GPU Compute Cores: 16384 Python Details - Python 3.10.8 Security Details - itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: 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: always-on RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected
rtx3090 arrayfire: Conjugate Gradient OpenCL blender: BMW27 - NVIDIA OptiX blender: Classroom - NVIDIA OptiX blender: Fishy Cat - NVIDIA OptiX blender: Barbershop - NVIDIA OptiX blender: Pabellon Barcelona - NVIDIA OptiX cl-mem: Copy cl-mem: Read cl-mem: Write clpeak: Integer Compute INT clpeak: Single-Precision Float clpeak: Double-Precision Double clpeak: Global Memory Bandwidth fahbench: financebench: Black-Scholes OpenCL hashcat: MD5 hashcat: SHA1 hashcat: 7-Zip hashcat: SHA-512 hashcat: TrueCrypt RIPEMD160 + XTS indigobench: OpenCL GPU - Bedroom indigobench: OpenCL GPU - Supercar lczero: OpenCL luxcorerender: DLSC - GPU luxcorerender: Danish Mood - GPU luxcorerender: Orange Juice - GPU luxcorerender: LuxCore Benchmark - GPU luxcorerender: Rainbow Colors and Prism - GPU mandelgpu: GPU namd-cuda: ATPase Simulation - 327,506 Atoms ncnn: Vulkan GPU - mobilenet ncnn: Vulkan GPU-v2-v2 - mobilenet-v2 ncnn: Vulkan GPU-v3-v3 - mobilenet-v3 ncnn: Vulkan GPU - shufflenet-v2 ncnn: Vulkan GPU - mnasnet ncnn: Vulkan GPU - efficientnet-b0 ncnn: Vulkan GPU - blazeface ncnn: Vulkan GPU - googlenet ncnn: Vulkan GPU - vgg16 ncnn: Vulkan GPU - resnet18 ncnn: Vulkan GPU - alexnet ncnn: Vulkan GPU - resnet50 ncnn: Vulkan GPU - yolov4-tiny ncnn: Vulkan GPU - squeezenet_ssd ncnn: Vulkan GPU - regnety_400m ncnn: Vulkan GPU - vision_transformer ncnn: Vulkan GPU - FastestDet neatbench: GPU octanebench: Total Score realsr-ncnn: 4x - No realsr-ncnn: 4x - Yes rodinia: OpenCL Particle Filter shoc: OpenCL - S3D shoc: OpenCL - Triad shoc: OpenCL - FFT SP shoc: OpenCL - MD5 Hash shoc: OpenCL - Reduction shoc: OpenCL - GEMM SGEMM_N shoc: OpenCL - Max SP Flops shoc: OpenCL - Bus Speed Download shoc: OpenCL - Bus Speed Readback shoc: OpenCL - Texture Read Bandwidth viennacl: CPU BLAS - sCOPY viennacl: CPU BLAS - sAXPY viennacl: CPU BLAS - sDOT viennacl: CPU BLAS - dCOPY viennacl: CPU BLAS - dAXPY viennacl: CPU BLAS - dDOT viennacl: CPU BLAS - dGEMV-N viennacl: CPU BLAS - dGEMV-T viennacl: CPU BLAS - dGEMM-NN viennacl: CPU BLAS - dGEMM-NT viennacl: CPU BLAS - dGEMM-TN viennacl: CPU BLAS - dGEMM-TT viennacl: OpenCL BLAS - sCOPY viennacl: OpenCL BLAS - sAXPY viennacl: OpenCL BLAS - sDOT viennacl: OpenCL BLAS - dCOPY viennacl: OpenCL BLAS - dAXPY viennacl: OpenCL BLAS - dDOT viennacl: OpenCL BLAS - dGEMV-N viennacl: OpenCL BLAS - dGEMV-T viennacl: OpenCL BLAS - dGEMM-NN viennacl: OpenCL BLAS - dGEMM-NT viennacl: OpenCL BLAS - dGEMM-TN viennacl: OpenCL BLAS - dGEMM-TT vkfft: vkpeak: fp32-scalar vkpeak: fp32-vec4 vkpeak: fp16-scalar vkpeak: fp16-vec4 vkpeak: fp64-scalar vkpeak: fp64-vec4 vkpeak: int32-scalar vkpeak: int32-vec4 vkpeak: int16-scalar vkpeak: int16-vec4 vkresample: 2x - Double vkresample: 2x - Single waifu2x-ncnn: 2x - 3 - Yes rtx3090-gpu rtx4090 1.598 5.78 14.00 10.15 51.61 15.43 362.3 827.1 751.5 18146.33 35529.52 653.69 814.22 328.4820 5.853 68180633333 21815700000 1092733 2735933333 790800 18.643 49.238 40156 11.28 9.17 10.18 11.14 31.03 3132109.6 0.09694 3.22 1.79 1.53 1.53 1.23 5.35 0.91 10.51 3.10 1.88 1.17 2.56 12.78 34.29 2.11 158.72 2.22 3090 686.766135 5.613 29.671 3.767 427.911 13.1025 2141.02 43.1315 395.854 8428.68 39391.9 13.3594 13.1913 2199.14 152 223 190 47.6 71.5 71.5 77.7 90.6 111 108 117 110 368 501 379 606 724 654 189 375 601 602 603 601 42747 20868.14 27457.96 20781.07 41153.13 652.97 652.88 20754.88 20661.07 13692.24 16740.59 118.989 9.350 3.263 0.8427 13.06 7.03 5.28 29.68 7.74 416.7 887.5 813.7 43083.60 85069.35 1449.77 869.72 442.4502 2.874 157233333333 51093900000 2775433 6520400000 1911500 33.701 77.657 74185 21.90 18.16 17.42 19.05 40.30 804490079.6 0.08095 6.98 2.66 2.93 2.89 2.50 3.78 0.85 2.32 3.37 1.62 2.14 3.75 9.46 8.16 1.48 270.65 3.72 4090 1366.29518 4.289 18.886 2.091 653.058 26.2868 2812.76 95.7989 1044.72 27907.7 92405.0 26.8302 26.3504 3098.67 162 245 218 48.5 72.4 73.0 81.5 98.8 111 108 116 110 482 610 460 664 776 732 225 450 1197 1320 1340 1390 63465 46187.77 61074.83 46166.64 91395.92 1455.07 1454.97 46167.66 45916.99 30756.40 40926.05 53.815 7.746 2.492 OpenBenchmarking.org
ArrayFire Test: Conjugate Gradient OpenCL OpenBenchmarking.org ms, Fewer Is Better ArrayFire 3.7 Test: Conjugate Gradient OpenCL rtx4090 rtx3090-gpu 0.3596 0.7192 1.0788 1.4384 1.798 SE +/- 0.0009, N = 3 SE +/- 0.0053, N = 3 0.8427 1.5980 1. (CXX) g++ options: -rdynamic
Blender Blend File: BMW27 - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.4 Blend File: BMW27 - Compute: NVIDIA OptiX rtx3090-gpu rtx4090 3 6 9 12 15 SE +/- 0.01, N = 3 SE +/- 9.70, N = 12 5.78 13.06
Blender Blend File: Classroom - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.4 Blend File: Classroom - Compute: NVIDIA OptiX rtx4090 rtx3090-gpu 4 8 12 16 20 SE +/- 0.02, N = 3 SE +/- 0.02, N = 3 7.03 14.00
Blender Blend File: Fishy Cat - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.4 Blend File: Fishy Cat - Compute: NVIDIA OptiX rtx4090 rtx3090-gpu 3 6 9 12 15 SE +/- 0.06, N = 13 SE +/- 0.01, N = 3 5.28 10.15
Blender Blend File: Barbershop - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.4 Blend File: Barbershop - Compute: NVIDIA OptiX rtx4090 rtx3090-gpu 12 24 36 48 60 SE +/- 0.06, N = 3 SE +/- 0.13, N = 3 29.68 51.61
Blender Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX OpenBenchmarking.org Seconds, Fewer Is Better Blender 3.4 Blend File: Pabellon Barcelona - Compute: NVIDIA OptiX rtx4090 rtx3090-gpu 4 8 12 16 20 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 7.74 15.43
cl-mem Benchmark: Copy OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Copy rtx4090 rtx3090-gpu 90 180 270 360 450 SE +/- 0.47, N = 3 SE +/- 0.35, N = 3 416.7 362.3 1. (CC) gcc options: -O2 -flto -lOpenCL
cl-mem Benchmark: Read OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Read rtx4090 rtx3090-gpu 200 400 600 800 1000 SE +/- 0.15, N = 3 SE +/- 0.59, N = 3 887.5 827.1 1. (CC) gcc options: -O2 -flto -lOpenCL
cl-mem Benchmark: Write OpenBenchmarking.org GB/s, More Is Better cl-mem 2017-01-13 Benchmark: Write rtx4090 rtx3090-gpu 200 400 600 800 1000 SE +/- 1.59, N = 3 SE +/- 0.58, N = 3 813.7 751.5 1. (CC) gcc options: -O2 -flto -lOpenCL
clpeak OpenCL Test: Integer Compute INT OpenBenchmarking.org GIOPS, More Is Better clpeak OpenCL Test: Integer Compute INT rtx4090 rtx3090-gpu 9K 18K 27K 36K 45K SE +/- 401.20, N = 3 SE +/- 152.55, N = 3 43083.60 18146.33 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
clpeak OpenCL Test: Single-Precision Float OpenBenchmarking.org GFLOPS, More Is Better clpeak OpenCL Test: Single-Precision Float rtx4090 rtx3090-gpu 20K 40K 60K 80K 100K SE +/- 31.18, N = 3 SE +/- 198.17, N = 3 85069.35 35529.52 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
clpeak OpenCL Test: Double-Precision Double OpenBenchmarking.org GFLOPS, More Is Better clpeak OpenCL Test: Double-Precision Double rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.83, N = 3 SE +/- 0.79, N = 3 1449.77 653.69 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
clpeak OpenCL Test: Global Memory Bandwidth OpenBenchmarking.org GBPS, More Is Better clpeak OpenCL Test: Global Memory Bandwidth rtx4090 rtx3090-gpu 200 400 600 800 1000 SE +/- 0.40, N = 3 SE +/- 0.03, N = 3 869.72 814.22 1. (CXX) g++ options: -O3 -rdynamic -lOpenCL
FAHBench OpenBenchmarking.org Ns Per Day, More Is Better FAHBench 2.3.2 rtx4090 rtx3090-gpu 100 200 300 400 500 SE +/- 0.52, N = 3 SE +/- 0.29, N = 3 442.45 328.48
FinanceBench Benchmark: Black-Scholes OpenCL OpenBenchmarking.org ms, Fewer Is Better FinanceBench 2016-07-25 Benchmark: Black-Scholes OpenCL rtx4090 rtx3090-gpu 1.3169 2.6338 3.9507 5.2676 6.5845 SE +/- 0.029, N = 6 SE +/- 0.004, N = 3 2.874 5.853 1. (CXX) g++ options: -O3 -march=native -fopenmp
Hashcat Benchmark: MD5 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: MD5 rtx4090 rtx3090-gpu 30000M 60000M 90000M 120000M 150000M SE +/- 120185042.52, N = 3 SE +/- 25348328.90, N = 3 157233333333 68180633333
Hashcat Benchmark: SHA1 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA1 rtx4090 rtx3090-gpu 11000M 22000M 33000M 44000M 55000M SE +/- 75227011.99, N = 3 SE +/- 45720163.31, N = 3 51093900000 21815700000
Hashcat Benchmark: 7-Zip OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: 7-Zip rtx4090 rtx3090-gpu 600K 1200K 1800K 2400K 3000K SE +/- 8294.64, N = 3 SE +/- 14591.13, N = 3 2775433 1092733
Hashcat Benchmark: SHA-512 OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: SHA-512 rtx4090 rtx3090-gpu 1400M 2800M 4200M 5600M 7000M SE +/- 2150193.79, N = 3 SE +/- 3640665.38, N = 3 6520400000 2735933333
Hashcat Benchmark: TrueCrypt RIPEMD160 + XTS OpenBenchmarking.org H/s, More Is Better Hashcat 6.2.4 Benchmark: TrueCrypt RIPEMD160 + XTS rtx4090 rtx3090-gpu 400K 800K 1200K 1600K 2000K SE +/- 18569.78, N = 6 SE +/- 8982.20, N = 3 1911500 790800
IndigoBench Acceleration: OpenCL GPU - Scene: Bedroom OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Bedroom rtx4090 rtx3090-gpu 8 16 24 32 40 SE +/- 0.06, N = 3 SE +/- 0.04, N = 3 33.70 18.64
IndigoBench Acceleration: OpenCL GPU - Scene: Supercar OpenBenchmarking.org M samples/s, More Is Better IndigoBench 4.4 Acceleration: OpenCL GPU - Scene: Supercar rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 0.02, N = 3 SE +/- 0.15, N = 3 77.66 49.24
LeelaChessZero Backend: OpenCL OpenBenchmarking.org Nodes Per Second, More Is Better LeelaChessZero 0.28 Backend: OpenCL rtx4090 rtx3090-gpu 16K 32K 48K 64K 80K SE +/- 538.71, N = 3 SE +/- 467.99, N = 3 74185 40156 1. (CXX) g++ options: -flto -pthread
LuxCoreRender Scene: DLSC - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: DLSC - Acceleration: GPU rtx4090 rtx3090-gpu 5 10 15 20 25 SE +/- 0.02, N = 3 SE +/- 0.01, N = 3 21.90 11.28 MIN: 19.4 / MAX: 22.16 MIN: 10.52 / MAX: 11.61
LuxCoreRender Scene: Danish Mood - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Danish Mood - Acceleration: GPU rtx4090 rtx3090-gpu 4 8 12 16 20 SE +/- 0.03, N = 3 SE +/- 0.02, N = 3 18.16 9.17 MIN: 7.14 / MAX: 21.2 MIN: 3.4 / MAX: 10.68
LuxCoreRender Scene: Orange Juice - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Orange Juice - Acceleration: GPU rtx4090 rtx3090-gpu 4 8 12 16 20 SE +/- 0.13, N = 3 SE +/- 0.02, N = 3 17.42 10.18 MIN: 13.8 / MAX: 24.29 MIN: 8.51 / MAX: 13.53
LuxCoreRender Scene: LuxCore Benchmark - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: LuxCore Benchmark - Acceleration: GPU rtx4090 rtx3090-gpu 5 10 15 20 25 SE +/- 0.03, N = 3 SE +/- 0.03, N = 3 19.05 11.14 MIN: 6.93 / MAX: 22.91 MIN: 3.22 / MAX: 12.94
LuxCoreRender Scene: Rainbow Colors and Prism - Acceleration: GPU OpenBenchmarking.org M samples/sec, More Is Better LuxCoreRender 2.6 Scene: Rainbow Colors and Prism - Acceleration: GPU rtx4090 rtx3090-gpu 9 18 27 36 45 SE +/- 0.26, N = 3 SE +/- 0.14, N = 3 40.30 31.03 MIN: 36.9 / MAX: 46 MIN: 29.8 / MAX: 34.64
MandelGPU OpenCL Device: GPU OpenBenchmarking.org Samples/sec, More Is Better MandelGPU 1.3pts1 OpenCL Device: GPU rtx4090 rtx3090-gpu 200M 400M 600M 800M 1000M SE +/- 5072564.87, N = 3 SE +/- 14.10, N = 3 804490079.6 3132109.6 1. (CC) gcc options: -O3 -lm -ftree-vectorize -funroll-loops -lglut -lOpenCL -lGL
NAMD CUDA ATPase Simulation - 327,506 Atoms OpenBenchmarking.org days/ns, Fewer Is Better NAMD CUDA 2.14 ATPase Simulation - 327,506 Atoms rtx4090 rtx3090-gpu 0.0218 0.0436 0.0654 0.0872 0.109 SE +/- 0.00092, N = 15 SE +/- 0.00139, N = 15 0.08095 0.09694
NCNN Target: Vulkan GPU - Model: mobilenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: mobilenet rtx3090-gpu rtx4090 2 4 6 8 10 SE +/- 0.03, N = 6 SE +/- 4.06, N = 9 3.22 6.98 MIN: 2.81 / MAX: 31.79 MIN: 2.61 / MAX: 47.34 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU-v2-v2 - Model: mobilenet-v2 rtx3090-gpu rtx4090 0.5985 1.197 1.7955 2.394 2.9925 SE +/- 0.03, N = 6 SE +/- 1.31, N = 9 1.79 2.66 MIN: 1.11 / MAX: 14.96 MIN: 1.16 / MAX: 20.75 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU-v3-v3 - Model: mobilenet-v3 rtx3090-gpu rtx4090 0.6593 1.3186 1.9779 2.6372 3.2965 SE +/- 0.03, N = 6 SE +/- 1.19, N = 9 1.53 2.93 MIN: 1.39 / MAX: 15.23 MIN: 1.25 / MAX: 41.37 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: shufflenet-v2 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: shufflenet-v2 rtx3090-gpu rtx4090 0.6503 1.3006 1.9509 2.6012 3.2515 SE +/- 0.05, N = 6 SE +/- 1.51, N = 9 1.53 2.89 MIN: 1.34 / MAX: 14.72 MIN: 1.17 / MAX: 21.26 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: mnasnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: mnasnet rtx3090-gpu rtx4090 0.5625 1.125 1.6875 2.25 2.8125 SE +/- 0.01, N = 6 SE +/- 1.41, N = 9 1.23 2.50 MIN: 1.13 / MAX: 7.82 MIN: 0.98 / MAX: 22.9 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: efficientnet-b0 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: efficientnet-b0 rtx4090 rtx3090-gpu 1.2038 2.4076 3.6114 4.8152 6.019 SE +/- 1.43, N = 9 SE +/- 0.96, N = 6 3.78 5.35 MIN: 1.76 / MAX: 23.24 MIN: 2.22 / MAX: 29.26 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: blazeface OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: blazeface rtx4090 rtx3090-gpu 0.2048 0.4096 0.6144 0.8192 1.024 SE +/- 0.00, N = 9 SE +/- 0.01, N = 6 0.85 0.91 MIN: 0.78 / MAX: 6.72 MIN: 0.83 / MAX: 8.02 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: googlenet OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: googlenet rtx4090 rtx3090-gpu 3 6 9 12 15 SE +/- 0.04, N = 9 SE +/- 0.10, N = 6 2.32 10.51 MIN: 1.45 / MAX: 14.91 MIN: 1.94 / MAX: 27.24 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: vgg16 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: vgg16 rtx3090-gpu rtx4090 0.7583 1.5166 2.2749 3.0332 3.7915 SE +/- 0.20, N = 6 SE +/- 1.30, N = 9 3.10 3.37 MIN: 1.59 / MAX: 25.34 MIN: 1.47 / MAX: 26.06 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: resnet18 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: resnet18 rtx4090 rtx3090-gpu 0.423 0.846 1.269 1.692 2.115 SE +/- 0.04, N = 7 SE +/- 0.11, N = 6 1.62 1.88 MIN: 1.17 / MAX: 14.28 MIN: 1.14 / MAX: 8.1 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: alexnet OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: alexnet rtx3090-gpu rtx4090 0.4815 0.963 1.4445 1.926 2.4075 SE +/- 0.01, N = 6 SE +/- 0.17, N = 9 1.17 2.14 MIN: 1.08 / MAX: 7.56 MIN: 1.02 / MAX: 20 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: resnet50 OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: resnet50 rtx3090-gpu rtx4090 0.8438 1.6876 2.5314 3.3752 4.219 SE +/- 0.09, N = 6 SE +/- 1.60, N = 9 2.56 3.75 MIN: 1.73 / MAX: 14.19 MIN: 1.46 / MAX: 25.78 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: yolov4-tiny OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: yolov4-tiny rtx4090 rtx3090-gpu 3 6 9 12 15 SE +/- 3.93, N = 9 SE +/- 3.53, N = 6 9.46 12.78 MIN: 4.95 / MAX: 48.71 MIN: 5.02 / MAX: 48.89 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: squeezenet_ssd OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: squeezenet_ssd rtx4090 rtx3090-gpu 8 16 24 32 40 SE +/- 4.27, N = 9 SE +/- 1.16, N = 6 8.16 34.29 MIN: 2.99 / MAX: 53.35 MIN: 7.52 / MAX: 75.03 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: regnety_400m OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: regnety_400m rtx4090 rtx3090-gpu 0.4748 0.9496 1.4244 1.8992 2.374 SE +/- 0.02, N = 9 SE +/- 0.17, N = 6 1.48 2.11 MIN: 1.35 / MAX: 14.94 MIN: 1.7 / MAX: 19.18 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: vision_transformer OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: vision_transformer rtx3090-gpu rtx4090 60 120 180 240 300 SE +/- 15.25, N = 6 SE +/- 127.68, N = 9 158.72 270.65 MIN: 115.97 / MAX: 1243.69 MIN: 99.36 / MAX: 1493.37 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NCNN Target: Vulkan GPU - Model: FastestDet OpenBenchmarking.org ms, Fewer Is Better NCNN 20220729 Target: Vulkan GPU - Model: FastestDet rtx3090-gpu rtx4090 0.837 1.674 2.511 3.348 4.185 SE +/- 0.06, N = 6 SE +/- 1.45, N = 9 2.22 3.72 MIN: 1.54 / MAX: 13.66 MIN: 1.65 / MAX: 21.61 1. (CXX) g++ options: -O3 -rdynamic -lgomp -lpthread
NeatBench Acceleration: GPU OpenBenchmarking.org FPS, More Is Better NeatBench 5 Acceleration: GPU rtx4090 rtx3090-gpu 900 1800 2700 3600 4500 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 4090 3090
OctaneBench Total Score OpenBenchmarking.org Score, More Is Better OctaneBench 2020.1 Total Score rtx4090 rtx3090-gpu 300 600 900 1200 1500 1366.30 686.77
RealSR-NCNN Scale: 4x - TAA: No OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: No rtx4090 rtx3090-gpu 1.2629 2.5258 3.7887 5.0516 6.3145 SE +/- 0.028, N = 3 SE +/- 0.011, N = 3 4.289 5.613
RealSR-NCNN Scale: 4x - TAA: Yes OpenBenchmarking.org Seconds, Fewer Is Better RealSR-NCNN 20200818 Scale: 4x - TAA: Yes rtx4090 rtx3090-gpu 7 14 21 28 35 SE +/- 0.01, N = 3 SE +/- 0.05, N = 3 18.89 29.67
Rodinia Test: OpenCL Particle Filter OpenBenchmarking.org Seconds, Fewer Is Better Rodinia 3.1 Test: OpenCL Particle Filter rtx4090 rtx3090-gpu 0.8476 1.6952 2.5428 3.3904 4.238 SE +/- 0.037, N = 6 2.091 3.767 1. (CXX) g++ options: -O2 -lOpenCL
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: S3D OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: S3D rtx4090 rtx3090-gpu 140 280 420 560 700 SE +/- 0.46, N = 3 SE +/- 0.53, N = 3 653.06 427.91 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Triad OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Triad rtx4090 rtx3090-gpu 6 12 18 24 30 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 26.29 13.10 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: FFT SP OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: FFT SP rtx4090 rtx3090-gpu 600 1200 1800 2400 3000 SE +/- 1.85, N = 3 SE +/- 0.44, N = 3 2812.76 2141.02 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: MD5 Hash OpenBenchmarking.org GHash/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: MD5 Hash rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 1.04, N = 4 SE +/- 0.06, N = 3 95.80 43.13 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Reduction OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Reduction rtx4090 rtx3090-gpu 200 400 600 800 1000 SE +/- 0.05, N = 3 SE +/- 0.54, N = 3 1044.72 395.85 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: GEMM SGEMM_N OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: GEMM SGEMM_N rtx4090 rtx3090-gpu 6K 12K 18K 24K 30K SE +/- 311.60, N = 4 SE +/- 58.26, N = 3 27907.70 8428.68 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Max SP Flops OpenBenchmarking.org GFLOPS, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Max SP Flops rtx4090 rtx3090-gpu 20K 40K 60K 80K 100K SE +/- 131.92, N = 3 SE +/- 204.68, N = 3 92405.0 39391.9 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Bus Speed Download OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Bus Speed Download rtx4090 rtx3090-gpu 6 12 18 24 30 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 26.83 13.36 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Bus Speed Readback OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Bus Speed Readback rtx4090 rtx3090-gpu 6 12 18 24 30 SE +/- 0.00, N = 3 SE +/- 0.00, N = 3 26.35 13.19 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
SHOC Scalable HeterOgeneous Computing Target: OpenCL - Benchmark: Texture Read Bandwidth OpenBenchmarking.org GB/s, More Is Better SHOC Scalable HeterOgeneous Computing 2020-04-17 Target: OpenCL - Benchmark: Texture Read Bandwidth rtx4090 rtx3090-gpu 700 1400 2100 2800 3500 SE +/- 1.63, N = 3 SE +/- 2.10, N = 3 3098.67 2199.14 1. (CXX) g++ options: -O2 -lSHOCCommonMPI -lSHOCCommonOpenCL -lSHOCCommon -lOpenCL -lrt -lmpi_cxx -lmpi
ViennaCL Test: CPU BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sCOPY rtx4090 rtx3090-gpu 40 80 120 160 200 SE +/- 1.86, N = 3 SE +/- 1.01, N = 15 162 152 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - sAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sAXPY rtx4090 rtx3090-gpu 50 100 150 200 250 SE +/- 5.13, N = 3 SE +/- 2.26, N = 15 245 223 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - sDOT rtx4090 rtx3090-gpu 50 100 150 200 250 SE +/- 6.66, N = 3 SE +/- 2.19, N = 15 218 190 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dCOPY rtx4090 rtx3090-gpu 11 22 33 44 55 SE +/- 0.23, N = 3 SE +/- 0.37, N = 15 48.5 47.6 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dAXPY rtx4090 rtx3090-gpu 16 32 48 64 80 SE +/- 1.57, N = 3 SE +/- 0.55, N = 15 72.4 71.5 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dDOT rtx4090 rtx3090-gpu 16 32 48 64 80 SE +/- 0.74, N = 3 SE +/- 0.81, N = 15 73.0 71.5 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-N rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 7.55, N = 3 SE +/- 4.91, N = 15 81.5 77.7 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMV-T rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 0.96, N = 3 SE +/- 2.55, N = 15 98.8 90.6 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NN rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 0.88, N = 3 SE +/- 0.48, N = 15 111 111 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-NT rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 0.33, N = 3 SE +/- 0.34, N = 15 108 108 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TN rtx3090-gpu rtx4090 30 60 90 120 150 SE +/- 0.59, N = 15 SE +/- 0.58, N = 3 117 116 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: CPU BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: CPU BLAS - dGEMM-TT rtx4090 rtx3090-gpu 20 40 60 80 100 SE +/- 0.58, N = 3 SE +/- 0.61, N = 10 110 110 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - sCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sCOPY rtx4090 rtx3090-gpu 100 200 300 400 500 SE +/- 0.67, N = 3 SE +/- 0.67, N = 3 482 368 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - sAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sAXPY rtx4090 rtx3090-gpu 130 260 390 520 650 SE +/- 0.58, N = 3 SE +/- 1.86, N = 3 610 501 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - sDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - sDOT rtx4090 rtx3090-gpu 100 200 300 400 500 SE +/- 0.33, N = 3 SE +/- 0.58, N = 3 460 379 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dCOPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dCOPY rtx4090 rtx3090-gpu 140 280 420 560 700 SE +/- 0.00, N = 3 SE +/- 0.58, N = 3 664 606 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dAXPY OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dAXPY rtx4090 rtx3090-gpu 200 400 600 800 1000 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 776 724 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dDOT OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dDOT rtx4090 rtx3090-gpu 160 320 480 640 800 SE +/- 0.58, N = 3 SE +/- 0.58, N = 3 732 654 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMV-N OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-N rtx4090 rtx3090-gpu 50 100 150 200 250 SE +/- 0.33, N = 3 SE +/- 0.33, N = 3 225 189 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMV-T OpenBenchmarking.org GB/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMV-T rtx4090 rtx3090-gpu 100 200 300 400 500 SE +/- 0.33, N = 3 SE +/- 0.00, N = 2 450 375 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-NN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NN rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 3.33, N = 3 SE +/- 1.86, N = 3 1197 601 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-NT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-NT rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.00, N = 3 SE +/- 4.41, N = 3 1320 602 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-TN OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TN rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.00, N = 3 SE +/- 1.53, N = 3 1340 603 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
ViennaCL Test: OpenCL BLAS - dGEMM-TT OpenBenchmarking.org GFLOPs/s, More Is Better ViennaCL 1.7.1 Test: OpenCL BLAS - dGEMM-TT rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.00, N = 3 1390 601 1. (CXX) g++ options: -fopenmp -O3 -rdynamic -lOpenCL
VkFFT OpenBenchmarking.org Benchmark Score, More Is Better VkFFT 1.1.1 rtx4090 rtx3090-gpu 14K 28K 42K 56K 70K SE +/- 1362.49, N = 6 SE +/- 394.43, N = 3 63465 42747 1. (CXX) g++ options: -O3
vkpeak fp32-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp32-scalar rtx4090 rtx3090-gpu 10K 20K 30K 40K 50K SE +/- 73.04, N = 3 SE +/- 62.27, N = 3 46187.77 20868.14
vkpeak fp32-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp32-vec4 rtx4090 rtx3090-gpu 13K 26K 39K 52K 65K SE +/- 47.86, N = 3 SE +/- 65.98, N = 3 61074.83 27457.96
vkpeak fp16-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp16-scalar rtx4090 rtx3090-gpu 10K 20K 30K 40K 50K SE +/- 46.75, N = 3 SE +/- 10.69, N = 3 46166.64 20781.07
vkpeak fp16-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp16-vec4 rtx4090 rtx3090-gpu 20K 40K 60K 80K 100K SE +/- 128.78, N = 3 SE +/- 12.43, N = 3 91395.92 41153.13
vkpeak fp64-scalar OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp64-scalar rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.06, N = 3 SE +/- 0.37, N = 3 1455.07 652.97
vkpeak fp64-vec4 OpenBenchmarking.org GFLOPS, More Is Better vkpeak 20210424 fp64-vec4 rtx4090 rtx3090-gpu 300 600 900 1200 1500 SE +/- 0.10, N = 3 SE +/- 0.37, N = 3 1454.97 652.88
vkpeak int32-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int32-scalar rtx4090 rtx3090-gpu 10K 20K 30K 40K 50K SE +/- 9.81, N = 3 SE +/- 9.80, N = 3 46167.66 20754.88
vkpeak int32-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int32-vec4 rtx4090 rtx3090-gpu 10K 20K 30K 40K 50K SE +/- 42.39, N = 3 SE +/- 6.94, N = 3 45916.99 20661.07
vkpeak int16-scalar OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int16-scalar rtx4090 rtx3090-gpu 7K 14K 21K 28K 35K SE +/- 26.05, N = 3 SE +/- 7.48, N = 3 30756.40 13692.24
vkpeak int16-vec4 OpenBenchmarking.org GIOPS, More Is Better vkpeak 20210424 int16-vec4 rtx4090 rtx3090-gpu 9K 18K 27K 36K 45K SE +/- 22.76, N = 3 SE +/- 2.73, N = 3 40926.05 16740.59
VkResample Upscale: 2x - Precision: Double OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Double rtx4090 rtx3090-gpu 30 60 90 120 150 SE +/- 0.02, N = 3 SE +/- 0.16, N = 3 53.82 118.99 1. (CXX) g++ options: -O3
VkResample Upscale: 2x - Precision: Single OpenBenchmarking.org ms, Fewer Is Better VkResample 1.0 Upscale: 2x - Precision: Single rtx4090 rtx3090-gpu 3 6 9 12 15 SE +/- 0.019, N = 3 SE +/- 0.063, N = 3 7.746 9.350 1. (CXX) g++ options: -O3
Waifu2x-NCNN Vulkan Scale: 2x - Denoise: 3 - TAA: Yes OpenBenchmarking.org Seconds, Fewer Is Better Waifu2x-NCNN Vulkan 20200818 Scale: 2x - Denoise: 3 - TAA: Yes rtx4090 rtx3090-gpu 0.7342 1.4684 2.2026 2.9368 3.671 SE +/- 0.009, N = 3 SE +/- 0.003, N = 3 2.492 3.263
Phoronix Test Suite v10.8.4