AMD EPYC 7601 32-Core testing with a ASRockRack EPYCD8 (P2.10 BIOS) and llvmpipe 126GB on Ubuntu 19.10 via the Phoronix Test Suite.
EPYC 7601 Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: ASRockRack EPYCD8 (P2.10 BIOS), Chipset: AMD 17h, Memory: 129024MB, Disk: 280GB INTEL SSDPED1D280GA, Graphics: llvmpipe 126GB, Network: 2 x Intel I350
OS: Ubuntu 19.10, Kernel: 5.4.0-999-generic (x86_64) 20191030, Desktop: GNOME Shell 3.34.1, Display Server: X Server 1.20.5, Display Driver: modesetting 1.20.5, OpenGL: 3.3 Mesa 19.2.1 (LLVM 9.0 128 bits), Compiler: GCC 9.2.1 20191008, File-System: ext4, Screen Resolution: 1024x768
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --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++,gm2 --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-offload-targets=nvptx-none,hsa --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-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: Scaling Governor: acpi-cpufreq ondemandSecurity Notes: 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 Full AMD retpoline IBPB: conditional STIBP: disabled RSB filling
EPYC 7601 Extra OpenBenchmarking.org Phoronix Test Suite AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads) ASRockRack EPYCD8 (P2.10 BIOS) AMD 17h 129024MB 280GB INTEL SSDPED1D280GA llvmpipe 126GB 2 x Intel I350 Ubuntu 19.10 5.4.0-999-generic (x86_64) 20191030 GNOME Shell 3.34.1 X Server 1.20.5 modesetting 1.20.5 3.3 Mesa 19.2.1 (LLVM 9.0 128 bits) GCC 9.2.1 20191008 ext4 1024x768 Processor Motherboard Chipset Memory Disk Graphics Network OS Kernel Desktop Display Server Display Driver OpenGL Compiler File-System Screen Resolution EPYC 7601 Extra Benchmarks System Logs - --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --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++,gm2 --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-offload-targets=nvptx-none,hsa --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-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 - Scaling Governor: acpi-cpufreq ondemand - 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 Full AMD retpoline IBPB: conditional STIBP: disabled RSB filling
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_3d - Data Type: f32 EPYC 7601 5 10 15 20 25 SE +/- 0.08, N = 3 18.67 MIN: 17.41 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 32 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 11.62, N = 3 6306.0 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 64 EPYC 7601 1200 2400 3600 4800 6000 SE +/- 3.36, N = 3 5728.3 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_all - Data Type: f32 EPYC 7601 600 1200 1800 2400 3000 SE +/- 2.32, N = 3 2566.44 MIN: 2437.38 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 32 EPYC 7601 2K 4K 6K 8K 10K SE +/- 1.38, N = 3 8158.4 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 64 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 6.55, N = 3 6452.5 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 128 EPYC 7601 1300 2600 3900 5200 6500 SE +/- 3.46, N = 3 6006.7 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_3d - Data Type: u8s8f32 EPYC 7601 1100 2200 3300 4400 5500 SE +/- 3.69, N = 3 4899.46 MIN: 4872.7 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 256 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 4.29, N = 3 6420.5 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 512 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 4.01, N = 3 6481.3 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 128 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 38.57, N = 3 6329.0 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 256 EPYC 7601 1300 2600 3900 5200 6500 SE +/- 5.21, N = 3 6141.8 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Deconvolution Batch deconv_3d - Data Type: f32 EPYC 7601 2 4 6 8 10 SE +/- 0.11045, N = 15 8.74631 MIN: 6.95 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 512 EPYC 7601 1300 2600 3900 5200 6500 SE +/- 12.86, N = 3 6114.1 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 1024 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 4.09, N = 3 6610.2 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 2048 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 5.48, N = 3 6388.6 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 1D FFT Size 4096 EPYC 7601 1400 2800 4200 5600 7000 SE +/- 5.98, N = 3 6316.3 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 1024 EPYC 7601 1200 2400 3600 4800 6000 SE +/- 6.38, N = 3 5521.0 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Deconvolution Batch deconv_all - Data Type: f32 EPYC 7601 700 1400 2100 2800 3500 SE +/- 23.55, N = 3 3129.28 MIN: 2970.16 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 2048 EPYC 7601 1100 2200 3300 4400 5500 SE +/- 25.21, N = 3 5065.2 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Stock - Size: 2D FFT Size 4096 EPYC 7601 1000 2000 3000 4000 5000 SE +/- 4.25, N = 3 4595.2 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 32 EPYC 7601 2K 4K 6K 8K 10K SE +/- 8.33, N = 3 10636 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 64 EPYC 7601 3K 6K 9K 12K 15K SE +/- 57.19, N = 3 13341 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 32 EPYC 7601 6K 12K 18K 24K 30K SE +/- 12.17, N = 3 26308 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 64 EPYC 7601 6K 12K 18K 24K 30K SE +/- 32.17, N = 3 26125 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Recurrent Neural Network Training - Data Type: f32 EPYC 7601 90 180 270 360 450 SE +/- 1.62, N = 3 413.24 MIN: 395.51 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 128 EPYC 7601 4K 8K 12K 16K 20K SE +/- 27.94, N = 3 16401 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: IP Batch 1D - Data Type: u8s8f32 EPYC 7601 12 24 36 48 60 SE +/- 0.18, N = 3 53.61 MIN: 46.59 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: IP Batch All - Data Type: u8s8f32 EPYC 7601 60 120 180 240 300 SE +/- 0.40, N = 3 272.38 MIN: 258.78 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: IP Batch 1D - Data Type: f32 EPYC 7601 1.1001 2.2002 3.3003 4.4004 5.5005 SE +/- 0.07959, N = 15 4.88919 MIN: 3.22 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: IP Batch All - Data Type: f32 EPYC 7601 7 14 21 28 35 SE +/- 0.43, N = 3 29.49 MIN: 27.7 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 256 EPYC 7601 4K 8K 12K 16K 20K SE +/- 137.11, N = 3 20976 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 512 EPYC 7601 5K 10K 15K 20K 25K SE +/- 85.41, N = 3 25663 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 128 EPYC 7601 5K 10K 15K 20K 25K SE +/- 42.81, N = 3 25492 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 256 EPYC 7601 5K 10K 15K 20K 25K SE +/- 45.35, N = 3 23334 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 512 EPYC 7601 5K 10K 15K 20K 25K SE +/- 25.75, N = 3 23576 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 1024 EPYC 7601 6K 12K 18K 24K 30K SE +/- 232.71, N = 3 27720 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Deconvolution Batch deconv_1d - Data Type: f32 EPYC 7601 2 4 6 8 10 SE +/- 0.07721, N = 6 6.52208 MIN: 5.59 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_googlenet_v3 - Data Type: u8s8f32 EPYC 7601 300 600 900 1200 1500 SE +/- 3.82, N = 3 1383.22 MIN: 1332.56 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 2048 EPYC 7601 6K 12K 18K 24K 30K SE +/- 284.85, N = 3 28028 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 1D FFT Size 4096 EPYC 7601 6K 12K 18K 24K 30K SE +/- 42.27, N = 3 27742 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_alexnet - Data Type: u8s8f32 EPYC 7601 500 1000 1500 2000 2500 SE +/- 5.12, N = 3 2480.57 MIN: 2430.35 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_googlenet_v3 - Data Type: f32 EPYC 7601 30 60 90 120 150 SE +/- 1.38, N = 3 147.57 MIN: 128.59 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Deconvolution Batch deconv_3d - Data Type: u8s8f32 EPYC 7601 700 1400 2100 2800 3500 SE +/- 5.06, N = 3 3138.11 MIN: 3124.28 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 1024 EPYC 7601 4K 8K 12K 16K 20K SE +/- 47.22, N = 3 20889 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Deconvolution Batch deconv_1d - Data Type: u8s8f32 EPYC 7601 400 800 1200 1600 2000 SE +/- 9.10, N = 3 1877.27 MIN: 1855.95 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 2048 EPYC 7601 3K 6K 9K 12K 15K SE +/- 29.63, N = 3 14876 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_all - Data Type: u8s8f32 EPYC 7601 5K 10K 15K 20K 25K SE +/- 59.53, N = 3 22469.1 MIN: 21801.7 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
FFTW FFTW is a C subroutine library for computing the discrete Fourier transform (DFT) in one or more dimensions. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Mflops, More Is Better FFTW 3.3.6 Build: Float + SSE - Size: 2D FFT Size 4096 EPYC 7601 3K 6K 9K 12K 15K SE +/- 46.89, N = 3 13423 1. (CC) gcc options: -pthread -O3 -fomit-frame-pointer -mtune=native -malign-double -fstrict-aliasing -fno-schedule-insns -ffast-math -lm
OpenBenchmarking.org FPS, More Is Better dav1d 0.5.0 Video Input: Summer Nature 4K EPYC 7601 30 60 90 120 150 SE +/- 0.45, N = 3 151.58 MIN: 94.31 / MAX: 164.18 1. (CC) gcc options: -pthread
MKL-DNN DNNL This is a test of the Intel MKL-DNN (DNNL / Deep Neural Network Library) as an Intel-optimized library for Deep Neural Networks and making use of its built-in benchdnn functionality. The result is the total perf time reported. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org ms, Fewer Is Better MKL-DNN DNNL 1.1 Harness: Convolution Batch conv_alexnet - Data Type: f32 EPYC 7601 70 140 210 280 350 SE +/- 1.22, N = 3 303.92 MIN: 291.88 1. (CXX) g++ options: -O3 -march=native -std=c++11 -msse4.1 -fPIC -fopenmp -pie -lpthread -ldl
OpenBenchmarking.org FPS, More Is Better dav1d 0.5.0 Video Input: Chimera 1080p 10-bit EPYC 7601 14 28 42 56 70 SE +/- 0.15, N = 3 61.55 MIN: 39.53 / MAX: 110.87 1. (CC) gcc options: -pthread
SVT-AV1 This is a test of the Intel Open Visual Cloud Scalable Video Technology SVT-AV1 CPU-based multi-threaded video encoder for the AV1 video format with a sample 1080p YUV video file. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 0.7 Encoder Mode: Enc Mode 0 - Input: 1080p EPYC 7601 0.0095 0.019 0.0285 0.038 0.0475 SE +/- 0.000, N = 6 0.042 1. (CXX) g++ options: -fPIE -fPIC -pie
OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 0.7 Encoder Mode: Enc Mode 4 - Input: 1080p EPYC 7601 0.92 1.84 2.76 3.68 4.6 SE +/- 0.037, N = 3 4.089 1. (CXX) g++ options: -fPIE -fPIC -pie
OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 0.7 Encoder Mode: Enc Mode 8 - Input: 1080p EPYC 7601 9 18 27 36 45 SE +/- 0.19, N = 3 40.19 1. (CXX) g++ options: -fPIE -fPIC -pie
OpenBenchmarking.org Frames Per Second, More Is Better Embree 3.6.1 Binary: Pathtracer ISPC - Model: Crown EPYC 7601 5 10 15 20 25 SE +/- 0.04, N = 3 19.54 MIN: 19.21 / MAX: 19.93
OpenBenchmarking.org Frames Per Second, More Is Better Embree 3.6.1 Binary: Pathtracer - Model: Asian Dragon EPYC 7601 6 12 18 24 30 SE +/- 0.03, N = 3 24.13 MIN: 23.93 / MAX: 24.48
OpenBenchmarking.org Frames Per Second, More Is Better Embree 3.6.1 Binary: Pathtracer - Model: Asian Dragon Obj EPYC 7601 5 10 15 20 25 SE +/- 0.01, N = 3 21.59 MIN: 21.41 / MAX: 21.92
OpenBenchmarking.org Frames Per Second, More Is Better Embree 3.6.1 Binary: Pathtracer ISPC - Model: Asian Dragon EPYC 7601 6 12 18 24 30 SE +/- 0.14, N = 3 23.11 MIN: 22.56 / MAX: 23.56
OpenBenchmarking.org Frames Per Second, More Is Better Embree 3.6.1 Binary: Pathtracer ISPC - Model: Asian Dragon Obj EPYC 7601 5 10 15 20 25 SE +/- 0.05, N = 3 20.13 MIN: 19.92 / MAX: 20.47
OSPray Intel OSPray is a portable ray-tracing engine for high-performance, high-fidenlity scientific visualizations. OSPray builds off Intel's Embree and Intel SPMD Program Compiler (ISPC) components as part of the oneAPI rendering toolkit. Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org FPS, More Is Better OSPray 1.8.5 Demo: San Miguel - Renderer: SciVis EPYC 7601 6 12 18 24 30 SE +/- 0.00, N = 12 24.39 MIN: 14.93 / MAX: 26.32
OpenBenchmarking.org FPS, More Is Better OSPray 1.8.5 Demo: XFrog Forest - Renderer: SciVis EPYC 7601 0.9765 1.953 2.9295 3.906 4.8825 SE +/- 0.01, N = 3 4.34 MIN: 3.95 / MAX: 4.41
OpenBenchmarking.org FPS, More Is Better OSPray 1.8.5 Demo: San Miguel - Renderer: Path Tracer EPYC 7601 0.4388 0.8776 1.3164 1.7552 2.194 SE +/- 0.00, N = 3 1.95 MIN: 1.54 / MAX: 1.98
OpenBenchmarking.org FPS, More Is Better OSPray 1.8.5 Demo: XFrog Forest - Renderer: Path Tracer EPYC 7601 0.513 1.026 1.539 2.052 2.565 SE +/- 0.00, N = 3 2.28 MIN: 2.11 / MAX: 2.31
EPYC 7601 Processor: AMD EPYC 7601 32-Core @ 2.20GHz (32 Cores / 64 Threads), Motherboard: ASRockRack EPYCD8 (P2.10 BIOS), Chipset: AMD 17h, Memory: 129024MB, Disk: 280GB INTEL SSDPED1D280GA, Graphics: llvmpipe 126GB, Network: 2 x Intel I350
OS: Ubuntu 19.10, Kernel: 5.4.0-999-generic (x86_64) 20191030, Desktop: GNOME Shell 3.34.1, Display Server: X Server 1.20.5, Display Driver: modesetting 1.20.5, OpenGL: 3.3 Mesa 19.2.1 (LLVM 9.0 128 bits), Compiler: GCC 9.2.1 20191008, File-System: ext4, Screen Resolution: 1024x768
Compiler Notes: --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --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++,gm2 --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-offload-targets=nvptx-none,hsa --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-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: Scaling Governor: acpi-cpufreq ondemandSecurity Notes: 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 Full AMD retpoline IBPB: conditional STIBP: disabled RSB filling
Testing initiated at 1 November 2019 22:28 by user phoronix.