AMD EPYC vs. Threadripper vs. Dual Xeon Gold

Some initial AMD EPYC 7601 tests on Ubuntu 17.04 with Linux 4.13. Tests for a future article on Phoronix.com. Benchmarks by Michael Larabel.

HTML result view exported from: https://openbenchmarking.org/result/1709106-TY-AMDEPYC2455&sor.

AMD EPYC vs. Threadripper vs. Dual Xeon GoldProcessorMotherboardChipsetMemoryDiskGraphicsMonitorNetworkAudioOSKernelDisplay DriverCompilerFile-SystemScreen ResolutionDesktopDisplay ServerOpenGLAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD Threadripper 1950XAMD EPYC 7601 32-Core @ 2.20GHz (64 Cores)TYAN B8026T70AE24HRAMD Device 1450129024MB234GBASPEED ASPEED FamilyAcer P243WBroadcom Limited NetXtreme BCM5720 Gigabit PCIeUbuntu 17.044.13.0-041300-generic (x86_64)modesetting 1.19.3GCC 6.3.0 20170406ext41920x1200Unity 7.5.02 x Intel Xeon Gold 6138 @ 3.70GHz (80 Cores)TYAN S7106Intel Device 202096256MB256GB Samsung SSD 850 + 2000GB Seagate ST2000DM006-2DM1 + 2 x 120GB TOSHIBA-TR150Intel I210 Gigabit ConnectionAMD Ryzen Threadripper 1950X 16-Core @ 3.40GHz (32 Cores)Gigabyte X399 AORUS Gaming 7AMD Device 145032768MB120GB Force MP500XFX AMD Radeon R9 290/390 4096MBRealtek ALC1220Acer B286HKQualcomm Atheros Device e0b1 + Intel Wireless 8265 / 8275X Server 1.19.34.5 Mesa 17.0.3 Gallium 0.4 (LLVM 4.0.0)3840x2160OpenBenchmarking.orgCompiler Details- --build=x86_64-linux-gnu --disable-browser-plugin --disable-vtable-verify --disable-werror --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-gtk-cairo --enable-java-awt=gtk --enable-java-home --enable-languages=c,ada,c++,java,go,d,fortran,objc,obj-c++ --enable-libmpx --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --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-arch-directory=amd64 --with-default-libstdcxx-abi=new --with-multilib-list=m32,m64,mx32 --with-target-system-zlib --with-tune=generic -v Processor Details- AMD EPYC 7601: Scaling Governor: acpi-cpufreq ondemand- AMD EPYC 7601 - NUMA Interleave All: Scaling Governor: acpi-cpufreq ondemand- 2 x Intel Xeon Gold 6138: Scaling Governor: intel_pstate powersave- AMD Threadripper 1950X: Scaling Governor: acpi-cpufreq ondemand

AMD EPYC vs. Threadripper vs. Dual Xeon Goldnpb: EP.Cnpb: FT.Bnpb: LU.Anpb: LU.Chpcc: G-HPLhpcc: G-Fftehpcc: G-Ptranshpcc: EP-STREAM Triadparboil: OpenMP LBMparboil: OpenMP Stencilrodinia: OpenMP LavaMDrodinia: OpenMP Streamclusterjohn-the-ripper: Blowfishttsiod-renderer: Phong Rendering With Soft-Shadow Mappingx264: H.264 Video Encodingbuild-linux-kernel: Time To Compilebuild-llvm: Time To Compilec-ray: Total Timeprimesieve: 1e12 Prime Number Generationopenssl: RSA 4096-bit Performancedarktable: Boat - CPU-onlydarktable: Masskrug - CPU-onlydarktable: Server Room - CPU-onlyblender: BMW27 - CPU-Onlyapache: Static Web Page Servingramspeed: Add - Integerhmmer: Pfam Database SearchAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD Threadripper 1950X1610.171660.0563042.8046983.344.643910.796500.783622.6541350.9513.8230.9623.1529553412.71292.6337.40175.112.8414.083294.537.949.244.62621.2922609.5832948.548.311607.682866.6662479.0850046.904.191300.535730.723803.5870338.377.7530.1414.6634335430.15288.2839.28192.262.8514.063306.377.037.573.72625.9322529.8333856.418.031815.892916.9953974.4950072.251.762670.925920.370354.3531550.047.7531.4422.5230373236.46310.7930.54134.642.8411.814826.7014.3013.145.29540.9828366.9620407.1016.311084.663398.8727592.4320949.026.013220.380010.540901.9958887.1013.1443.7137.6314024424.77324.4945.96249.084.2220.262189.7011.6010.826.22525.4127960.0633915.397.77OpenBenchmarking.org

NAS Parallel Benchmarks

Test / Class: EP.C

OpenBenchmarking.orgTotal Mop/s, More Is BetterNAS Parallel Benchmarks 3.3Test / Class: EP.C2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X400800120016002000SE +/- 34.76, N = 6SE +/- 0.32, N = 3SE +/- 0.90, N = 3SE +/- 0.58, N = 31815.891610.171607.681084.661. (F9X) gfortran options: -O3 -march=native -pthread -lmpi_usempif08 -lmpi_mpifh -lmpi2. Open MPI 2.0.2

NAS Parallel Benchmarks

Test / Class: FT.B

OpenBenchmarking.orgTotal Mop/s, More Is BetterNAS Parallel Benchmarks 3.3Test / Class: FT.BAMD Threadripper 1950X2 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD EPYC 76017001400210028003500SE +/- 2.42, N = 3SE +/- 42.85, N = 4SE +/- 2.64, N = 3SE +/- 44.23, N = 63398.872916.992866.661660.051. (F9X) gfortran options: -O3 -march=native -pthread -lmpi_usempif08 -lmpi_mpifh -lmpi2. Open MPI 2.0.2

NAS Parallel Benchmarks

Test / Class: LU.A

OpenBenchmarking.orgTotal Mop/s, More Is BetterNAS Parallel Benchmarks 3.3Test / Class: LU.AAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD Threadripper 1950X14K28K42K56K70KSE +/- 1051.01, N = 4SE +/- 437.89, N = 3SE +/- 3783.72, N = 6SE +/- 11.08, N = 363042.8062479.0853974.4927592.431. (F9X) gfortran options: -O3 -march=native -pthread -lmpi_usempif08 -lmpi_mpifh -lmpi2. Open MPI 2.0.2

NAS Parallel Benchmarks

Test / Class: LU.C

OpenBenchmarking.orgTotal Mop/s, More Is BetterNAS Parallel Benchmarks 3.3Test / Class: LU.C2 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X11K22K33K44K55KSE +/- 637.78, N = 3SE +/- 25.06, N = 3SE +/- 747.98, N = 3SE +/- 91.81, N = 350072.2550046.9046983.3420949.021. (F9X) gfortran options: -O3 -march=native -pthread -lmpi_usempif08 -lmpi_mpifh -lmpi2. Open MPI 2.0.2

HPC Challenge

Test / Class: G-HPL

OpenBenchmarking.orgGFLOPS, More Is BetterHPC Challenge 1.5.0Test / Class: G-HPLAMD Threadripper 1950XAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138246810SE +/- 0.17770, N = 6SE +/- 0.06846, N = 3SE +/- 0.05229, N = 3SE +/- 0.02653, N = 36.013224.643914.191301.762671. (CC) gcc options: -lblas -lm -lmpich2. BLAS + mpicc for MPICH version 3.2

HPC Challenge

Test / Class: G-Ffte

OpenBenchmarking.orgGFLOPS, More Is BetterHPC Challenge 1.5.0Test / Class: G-Ffte2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X0.20830.41660.62490.83321.0415SE +/- 0.10042, N = 3SE +/- 0.00574, N = 3SE +/- 0.08554, N = 3SE +/- 0.02518, N = 30.925920.796500.535730.380011. (CC) gcc options: -lblas -lm -lmpich2. BLAS + mpicc for MPICH version 3.2

HPC Challenge

Test / Class: G-Ffte

OpenBenchmarking.orgGFLOP/s, More Is BetterHPC Challenge 1.5.0Test / Class: G-Ffte2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X0.20830.41660.62490.83321.0415SE +/- 0.10042, N = 3SE +/- 0.00574, N = 3SE +/- 0.08554, N = 3SE +/- 0.02518, N = 30.925920.796500.535730.380011. (CC) gcc options: -lblas -lm -lmpich2. BLAS + mpicc for MPICH version 3.2

HPC Challenge

Test / Class: G-Ptrans

OpenBenchmarking.orgGB/s, More Is BetterHPC Challenge 1.5.0Test / Class: G-PtransAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X2 x Intel Xeon Gold 61380.17630.35260.52890.70520.8815SE +/- 0.02928, N = 3SE +/- 0.06341, N = 3SE +/- 0.05016, N = 3SE +/- 0.04259, N = 30.783620.723800.540900.370351. (CC) gcc options: -lblas -lm -lmpich2. BLAS + mpicc for MPICH version 3.2

HPC Challenge

Test / Class: EP-STREAM Triad

OpenBenchmarking.orgGB/s, More Is BetterHPC Challenge 1.5.0Test / Class: EP-STREAM Triad2 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X0.97951.9592.93853.9184.8975SE +/- 0.84575, N = 3SE +/- 1.07962, N = 3SE +/- 0.11259, N = 3SE +/- 0.04876, N = 34.353153.587032.654131.995881. (CC) gcc options: -lblas -lm -lmpich2. BLAS + mpicc for MPICH version 3.2

Parboil

Test: OpenMP LBM

OpenBenchmarking.orgSeconds, Fewer Is BetterParboil 2.5Test: OpenMP LBMAMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD EPYC 7601AMD Threadripper 1950X20406080100SE +/- 0.26, N = 3SE +/- 1.14, N = 6SE +/- 0.66, N = 3SE +/- 0.05, N = 338.3750.0450.9587.101. (CXX) g++ options: -lm -lpthread -lgomp -ffast-math -fopenmp

Parboil

Test: OpenMP Stencil

OpenBenchmarking.orgSeconds, Fewer Is BetterParboil 2.5Test: OpenMP StencilAMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD Threadripper 1950XAMD EPYC 760148121620SE +/- 0.00, N = 3SE +/- 0.17, N = 6SE +/- 0.08, N = 3SE +/- 0.68, N = 67.757.7513.1413.821. (CXX) g++ options: -lm -lpthread -lgomp -ffast-math -fopenmp

Rodinia

Test: OpenMP LavaMD

OpenBenchmarking.orgSeconds, Fewer Is BetterRodinia 2.4Test: OpenMP LavaMDAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 76012 x Intel Xeon Gold 6138AMD Threadripper 1950X1020304050SE +/- 0.15, N = 3SE +/- 0.01, N = 3SE +/- 0.10, N = 3SE +/- 0.07, N = 330.1430.9631.4443.711. (CXX) g++ options: -O2 -lOpenCL

Rodinia

Test: OpenMP Streamcluster

OpenBenchmarking.orgSeconds, Fewer Is BetterRodinia 2.4Test: OpenMP StreamclusterAMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD EPYC 7601AMD Threadripper 1950X918273645SE +/- 0.27, N = 6SE +/- 0.42, N = 3SE +/- 1.43, N = 6SE +/- 0.74, N = 314.6622.5223.1537.631. (CXX) g++ options: -O2 -lOpenCL

John The Ripper

Test: Blowfish

OpenBenchmarking.orgReal C/S, More Is BetterJohn The Ripper 1.8.0Test: BlowfishAMD EPYC 7601 - NUMA Interleave All2 x Intel Xeon Gold 6138AMD EPYC 7601AMD Threadripper 1950X7K14K21K28K35KSE +/- 764.17, N = 6SE +/- 2076.33, N = 6SE +/- 2253.07, N = 6SE +/- 2257.07, N = 6343353037329553140241. (CC) gcc options: -fopenmp -lcrypt

TTSIOD 3D Renderer

Phong Rendering With Soft-Shadow Mapping

OpenBenchmarking.orgFPS, More Is BetterTTSIOD 3D Renderer 2.3aPhong Rendering With Soft-Shadow MappingAMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950XAMD EPYC 76012 x Intel Xeon Gold 613890180270360450SE +/- 3.57, N = 3SE +/- 3.24, N = 3SE +/- 4.72, N = 3SE +/- 17.30, N = 6430.15424.77412.71236.461. (CXX) g++ options: -O3 -fomit-frame-pointer -ffast-math -mtune=native -flto -msse -mrecip -mfpmath=sse -msse2 -mssse3 -lSDL -lstdc++

x264

H.264 Video Encoding

OpenBenchmarking.orgFrames Per Second, More Is Betterx264 2017-09-08H.264 Video EncodingAMD Threadripper 1950X2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All70140210280350SE +/- 1.53, N = 3SE +/- 3.63, N = 3SE +/- 1.27, N = 3SE +/- 0.45, N = 3324.49310.79292.63288.28-lavformat -lavcodec -lavutil -lswscale-lavformat -lavcodec -lavutil -lswscale1. (CC) gcc options: -ldl -m64 -lm -lpthread -O3 -ffast-math -std=gnu99 -fPIC -fomit-frame-pointer -fno-tree-vectorize

Timed Linux Kernel Compilation

Time To Compile

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed Linux Kernel Compilation 4.9Time To Compile2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X1020304050SE +/- 0.89, N = 6SE +/- 0.51, N = 6SE +/- 0.58, N = 5SE +/- 0.62, N = 330.5437.4039.2845.96

Timed LLVM Compilation

Time To Compile

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed LLVM Compilation 4.0.1Time To Compile2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X50100150200250SE +/- 0.81, N = 3SE +/- 3.02, N = 4SE +/- 2.34, N = 3SE +/- 3.05, N = 3134.64175.11192.26249.08

C-Ray

Total Time

OpenBenchmarking.orgSeconds, Fewer Is BetterC-Ray 1.1Total TimeAMD EPYC 76012 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD Threadripper 1950X0.94951.8992.84853.7984.7475SE +/- 0.02, N = 3SE +/- 0.01, N = 3SE +/- 0.03, N = 3SE +/- 0.02, N = 32.842.842.854.221. (CC) gcc options: -lm -lpthread -O3

Primesieve

1e12 Prime Number Generation

OpenBenchmarking.orgSeconds, Fewer Is BetterPrimesieve 5.4.21e12 Prime Number Generation2 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X510152025SE +/- 0.03, N = 3SE +/- 0.04, N = 3SE +/- 0.05, N = 3SE +/- 0.01, N = 311.8114.0614.0820.261. (CXX) g++ options: -O2 -fopenmp

OpenSSL

RSA 4096-bit Performance

OpenBenchmarking.orgSigns Per Second, More Is BetterOpenSSL 1.0.1gRSA 4096-bit Performance2 x Intel Xeon Gold 6138AMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X10002000300040005000SE +/- 23.22, N = 3SE +/- 13.22, N = 3SE +/- 14.45, N = 3SE +/- 1.96, N = 34826.703306.373294.532189.701. (CC) gcc options: -m64 -O3 -lssl -lcrypto -ldl

Darktable

Test: Boat - Acceleration: CPU-only

OpenBenchmarking.orgSeconds, Fewer Is BetterDarktable 2.2.1Test: Boat - Acceleration: CPU-onlyAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X2 x Intel Xeon Gold 613848121620SE +/- 0.05, N = 3SE +/- 0.05, N = 3SE +/- 0.10, N = 3SE +/- 0.59, N = 67.037.9411.6014.30

Darktable

Test: Masskrug - Acceleration: CPU-only

OpenBenchmarking.orgSeconds, Fewer Is BetterDarktable 2.2.1Test: Masskrug - Acceleration: CPU-onlyAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 7601AMD Threadripper 1950X2 x Intel Xeon Gold 61383691215SE +/- 0.13, N = 4SE +/- 0.09, N = 3SE +/- 0.07, N = 3SE +/- 0.27, N = 67.579.2410.8213.14

Darktable

Test: Server Room - Acceleration: CPU-only

OpenBenchmarking.orgSeconds, Fewer Is BetterDarktable 2.2.1Test: Server Room - Acceleration: CPU-onlyAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 76012 x Intel Xeon Gold 6138AMD Threadripper 1950X246810SE +/- 0.08, N = 6SE +/- 0.07, N = 6SE +/- 0.35, N = 6SE +/- 0.09, N = 33.724.625.296.22

Blender

Blend File: BMW27 - Compute: CPU-Only

OpenBenchmarking.orgSeconds, Fewer Is BetterBlender 2.78cBlend File: BMW27 - Compute: CPU-OnlyAMD Threadripper 1950X2 x Intel Xeon Gold 6138AMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All140280420560700525.41540.98621.29625.93

Apache Benchmark

Static Web Page Serving

OpenBenchmarking.orgRequests Per Second, More Is BetterApache Benchmark 2.4.7Static Web Page Serving2 x Intel Xeon Gold 6138AMD Threadripper 1950XAMD EPYC 7601AMD EPYC 7601 - NUMA Interleave All6K12K18K24K30KSE +/- 105.14, N = 3SE +/- 168.94, N = 3SE +/- 50.73, N = 3SE +/- 52.07, N = 328366.9627960.0622609.5822529.831. (CC) gcc options: -shared -fPIC -O2 -pthread

RAMspeed SMP

Type: Add - Benchmark: Integer

OpenBenchmarking.orgMB/s, More Is BetterRAMspeed SMP 3.5.0Type: Add - Benchmark: IntegerAMD Threadripper 1950XAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 76012 x Intel Xeon Gold 61387K14K21K28K35K33915.3933856.4132948.5420407.101. (CC) gcc options: -O3 -march=native

Timed HMMer Search

Pfam Database Search

OpenBenchmarking.orgSeconds, Fewer Is BetterTimed HMMer Search 2.3.2Pfam Database SearchAMD Threadripper 1950XAMD EPYC 7601 - NUMA Interleave AllAMD EPYC 76012 x Intel Xeon Gold 613848121620SE +/- 0.05, N = 3SE +/- 0.03, N = 3SE +/- 0.05, N = 3SE +/- 0.08, N = 37.778.038.3116.311. (CC) gcc options: -O2 -pthread -lhmmer -lsquid -lm

Phoronix Test Suite v10.8.4