new ai

AMD Ryzen 9 7950X 16-Core testing with a ASUS ROG CROSSHAIR X670E HERO (1101 BIOS) and AMD Radeon RX 7900 XTX 24GB on Ubuntu 22.04 via the Phoronix Test Suite.

Compare your own system(s) to this result file with the Phoronix Test Suite by running the command: phoronix-test-suite benchmark 2304292-PTS-NEWAI12536
Jump To Table - Results

View

Do Not Show Noisy Results
Do Not Show Results With Incomplete Data
Do Not Show Results With Little Change/Spread
List Notable Results

Statistics

Show Overall Harmonic Mean(s)
Show Overall Geometric Mean
Show Wins / Losses Counts (Pie Chart)
Normalize Results
Remove Outliers Before Calculating Averages

Graph Settings

Force Line Graphs Where Applicable
Convert To Scalar Where Applicable
Prefer Vertical Bar Graphs

Multi-Way Comparison

Condense Multi-Option Tests Into Single Result Graphs

Table

Show Detailed System Result Table

Run Management

Highlight
Result
Hide
Result
Result
Identifier
Performance Per
Dollar
Date
Run
  Test
  Duration
a
April 28 2023
  7 Hours, 6 Minutes
b
April 28 2023
  3 Hours, 42 Minutes
c
April 28 2023
  7 Hours, 11 Minutes
Invert Hiding All Results Option
  6 Hours

Only show results where is faster than
Only show results matching title/arguments (delimit multiple options with a comma):
Do not show results matching title/arguments (delimit multiple options with a comma):

a

Kernel Notes: Transparent Huge Pages: madvise
Compiler 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-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 Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203
Python Notes: Python 3.10.9
Security Notes: 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 + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

b

c

Processor: AMD Ryzen 9 7950X 16-Core @ 4.50GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR X670E HERO (1101 BIOS), Chipset: AMD Device 14d8, Memory: 32GB, Disk: 2048GB SOLIDIGM SSDPFKKW020X7 + 2000GB, Graphics: AMD Radeon RX 7900 XTX 24GB (2304/1249MHz), Audio: AMD Device ab30, Monitor: ASUS MG28U, Network: Intel I225-V + Intel Wi-Fi 6 AX210/AX211/AX411

OS: Ubuntu 22.04, Kernel: 6.3.0-060300rc7daily20230417-generic (x86_64), Desktop: GNOME Shell 42.5, Display Server: X Server 1.21.1.3 + Wayland, OpenGL: 4.6 Mesa 23.2.0-devel (git-f6fb189 2023-04-18 jammy-oibaf-ppa) (LLVM 15.0.7 DRM 3.52), Vulkan: 1.3.246, Compiler: GCC 11.3.0, File-System: ext4, Screen Resolution: 3840x2160

new aiOpenBenchmarking.orgPhoronix Test SuiteAMD Ryzen 9 7950X 16-Core @ 4.50GHz (16 Cores / 32 Threads)ASUS ROG CROSSHAIR X670E HERO (1101 BIOS)AMD Device 14d832GB2048GB SOLIDIGM SSDPFKKW020X7 + 2000GBAMD Radeon RX 7900 XTX 24GB (2304/1249MHz)AMD Device ab30ASUS MG28UIntel I225-V + Intel Wi-Fi 6 AX210/AX211/AX411Ubuntu 22.046.3.0-060300rc7daily20230417-generic (x86_64)GNOME Shell 42.5X Server 1.21.1.3 + Wayland4.6 Mesa 23.2.0-devel (git-f6fb189 2023-04-18 jammy-oibaf-ppa) (LLVM 15.0.7 DRM 3.52)1.3.246GCC 11.3.0ext43840x2160ProcessorMotherboardChipsetMemoryDiskGraphicsAudioMonitorNetworkOSKernelDesktopDisplay ServerOpenGLVulkanCompilerFile-SystemScreen ResolutionNew Ai BenchmarksSystem Logs- Transparent Huge Pages: madvise- --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 - Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203- Python 3.10.9- 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 + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

abcResult OverviewPhoronix Test Suite100%100%100%100%Intel TensorFlowSVT-AV1Faiss

new aisvt-av1: Preset 4 - Bosphorus 4Ksvt-av1: Preset 8 - Bosphorus 4Ksvt-av1: Preset 12 - Bosphorus 4Ksvt-av1: Preset 13 - Bosphorus 4Ksvt-av1: Preset 4 - Bosphorus 1080psvt-av1: Preset 8 - Bosphorus 1080psvt-av1: Preset 12 - Bosphorus 1080psvt-av1: Preset 13 - Bosphorus 1080pfaiss: demo_sift1Mfaiss: bench_polysemous_sift1m - PQ baselinefaiss: bench_polysemous_sift1m - Polysemous 64faiss: bench_polysemous_sift1m - Polysemous 62faiss: bench_polysemous_sift1m - Polysemous 58faiss: bench_polysemous_sift1m - Polysemous 54faiss: bench_polysemous_sift1m - Polysemous 50faiss: bench_polysemous_sift1m - Polysemous 46faiss: bench_polysemous_sift1m - Polysemous 42faiss: bench_polysemous_sift1m - Polysemous 38faiss: bench_polysemous_sift1m - Polysemous 34faiss: bench_polysemous_sift1m - Polysemous 30intel-tensorflow: resnet50_fp32_pretrained_model - 1intel-tensorflow: resnet50_fp32_pretrained_model - 1intel-tensorflow: resnet50_int8_pretrained_model - 1intel-tensorflow: resnet50_int8_pretrained_model - 1intel-tensorflow: resnet50_fp32_pretrained_model - 16intel-tensorflow: resnet50_fp32_pretrained_model - 32intel-tensorflow: resnet50_fp32_pretrained_model - 64intel-tensorflow: resnet50_fp32_pretrained_model - 96intel-tensorflow: resnet50_int8_pretrained_model - 16intel-tensorflow: resnet50_int8_pretrained_model - 32intel-tensorflow: resnet50_int8_pretrained_model - 64intel-tensorflow: resnet50_int8_pretrained_model - 96intel-tensorflow: resnet50_fp32_pretrained_model - 256intel-tensorflow: resnet50_fp32_pretrained_model - 512intel-tensorflow: resnet50_int8_pretrained_model - 256intel-tensorflow: resnet50_int8_pretrained_model - 512intel-tensorflow: inceptionv4_fp32_pretrained_model - 1intel-tensorflow: inceptionv4_fp32_pretrained_model - 1intel-tensorflow: inceptionv4_int8_pretrained_model - 1intel-tensorflow: inceptionv4_int8_pretrained_model - 1intel-tensorflow: mobilenetv1_fp32_pretrained_model - 1intel-tensorflow: mobilenetv1_int8_pretrained_model - 1intel-tensorflow: inceptionv4_fp32_pretrained_model - 16intel-tensorflow: inceptionv4_fp32_pretrained_model - 32intel-tensorflow: inceptionv4_fp32_pretrained_model - 64intel-tensorflow: inceptionv4_fp32_pretrained_model - 96intel-tensorflow: inceptionv4_int8_pretrained_model - 16intel-tensorflow: inceptionv4_int8_pretrained_model - 32intel-tensorflow: inceptionv4_int8_pretrained_model - 64intel-tensorflow: inceptionv4_int8_pretrained_model - 96intel-tensorflow: mobilenetv1_fp32_pretrained_model - 16intel-tensorflow: mobilenetv1_fp32_pretrained_model - 32intel-tensorflow: mobilenetv1_fp32_pretrained_model - 64intel-tensorflow: mobilenetv1_fp32_pretrained_model - 96intel-tensorflow: mobilenetv1_int8_pretrained_model - 16intel-tensorflow: mobilenetv1_int8_pretrained_model - 32intel-tensorflow: mobilenetv1_int8_pretrained_model - 64intel-tensorflow: mobilenetv1_int8_pretrained_model - 96intel-tensorflow: inceptionv4_fp32_pretrained_model - 256intel-tensorflow: inceptionv4_fp32_pretrained_model - 512intel-tensorflow: inceptionv4_int8_pretrained_model - 256intel-tensorflow: inceptionv4_int8_pretrained_model - 512intel-tensorflow: mobilenetv1_fp32_pretrained_model - 256intel-tensorflow: mobilenetv1_fp32_pretrained_model - 512intel-tensorflow: mobilenetv1_int8_pretrained_model - 256intel-tensorflow: mobilenetv1_int8_pretrained_model - 512zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 1zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 1zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 16zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 32zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 64zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 96zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 16zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 32zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 64zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 96zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 256zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 512zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 256zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 512zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 1zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 16zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 32zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 64zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 96zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 256zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 512abc6.11175.893210.908206.52014.168119.996716.460698.56764.0092.6244.1273.4662.1321.2840.8300.6120.5250.4950.4850.482139.0517.192547.7771.826264.850252.894247.813245.027800.103824.027830.205803.346247.463249.615779.333774.17761.6517.070104.9610.5601591.734525.9083.6380.4279.1579.05311.78312.25309.86300.911107.681047.32952.20914.914671.255094.814327.784293.1080.3681.04288.45290.66867.02856.503482.013263.948.07023.159241.34228.35225.81228.1162.7862.8264.6365.86238.11242.6568.2470.361.586953.78770.81715.78691.01650.17643.446.16575.807213.399207.37314.091120.306711.29699.65263.9032.6134.0913.4492.1261.2780.8250.6100.5220.4940.4850.481139.5197.168503.5001.986267.531253.600247.959246.187798.920825.337837.083803.687247.089249.918779.241775.25861.0017.296105.5010.3511573.544539.9083.2580.5679.5079.44319.25315.79309.26303.131108.831047.98951.88915.834689.945078.874330.684285.7680.4081.07289.43288.25867.136.11175.612210.656206.94314.234118.000710.846698.00964.0602.6274.1173.4682.1341.2850.8310.6120.5250.4960.4860.483135.8947.360523.1711.924267.660252.684248.704245.605793.970830.170833.729805.521247.448249.763780.454775.55757.4217.20099.0010.4451606.414528.3282.6580.2879.0078.91312.28311.68314.14297.391107.451048.20951.85915.294667.755092.764334.434278.5880.2080.95287.48288.82866.92855.613482.913272.668.12223.131240.72226.86226.03230.1462.7962.9164.6165.59239.23243.2968.4270.061.573953.58768.31717.29690.86650.04642.22OpenBenchmarking.org

SVT-AV1

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 4 - Input: Bosphorus 4Kabc246810SE +/- 0.030, N = 3SE +/- 0.017, N = 3SE +/- 0.004, N = 36.1116.1656.1111. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 4 - Input: Bosphorus 4Kabc246810Min: 6.05 / Avg: 6.11 / Max: 6.15Min: 6.14 / Avg: 6.16 / Max: 6.19Min: 6.11 / Avg: 6.11 / Max: 6.121. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 8 - Input: Bosphorus 4Kabc20406080100SE +/- 0.39, N = 3SE +/- 0.11, N = 3SE +/- 0.56, N = 375.8975.8175.611. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 8 - Input: Bosphorus 4Kabc1530456075Min: 75.19 / Avg: 75.89 / Max: 76.55Min: 75.69 / Avg: 75.81 / Max: 76.02Min: 74.66 / Avg: 75.61 / Max: 76.611. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 12 - Input: Bosphorus 4Kabc50100150200250SE +/- 1.35, N = 3SE +/- 0.58, N = 3SE +/- 0.55, N = 3210.91213.40210.661. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 12 - Input: Bosphorus 4Kabc4080120160200Min: 208.2 / Avg: 210.91 / Max: 212.35Min: 212.33 / Avg: 213.4 / Max: 214.33Min: 209.99 / Avg: 210.66 / Max: 211.741. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 13 - Input: Bosphorus 4Kabc50100150200250SE +/- 1.23, N = 3SE +/- 0.95, N = 3SE +/- 0.70, N = 3206.52207.37206.941. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 13 - Input: Bosphorus 4Kabc4080120160200Min: 204.07 / Avg: 206.52 / Max: 207.86Min: 205.87 / Avg: 207.37 / Max: 209.13Min: 205.65 / Avg: 206.94 / Max: 208.031. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 4 - Input: Bosphorus 1080pabc48121620SE +/- 0.02, N = 3SE +/- 0.08, N = 3SE +/- 0.05, N = 314.1714.0914.231. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 4 - Input: Bosphorus 1080pabc48121620Min: 14.14 / Avg: 14.17 / Max: 14.21Min: 13.97 / Avg: 14.09 / Max: 14.26Min: 14.14 / Avg: 14.23 / Max: 14.291. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 8 - Input: Bosphorus 1080pabc306090120150SE +/- 0.60, N = 3SE +/- 0.75, N = 3SE +/- 0.63, N = 3120.00120.31118.001. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 8 - Input: Bosphorus 1080pabc20406080100Min: 118.8 / Avg: 120 / Max: 120.68Min: 118.95 / Avg: 120.31 / Max: 121.55Min: 117.32 / Avg: 118 / Max: 119.261. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 12 - Input: Bosphorus 1080pabc150300450600750SE +/- 5.15, N = 3SE +/- 8.08, N = 3SE +/- 6.31, N = 3716.46711.29710.851. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 12 - Input: Bosphorus 1080pabc130260390520650Min: 708.02 / Avg: 716.46 / Max: 725.78Min: 696.57 / Avg: 711.29 / Max: 724.44Min: 698.25 / Avg: 710.85 / Max: 717.921. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 13 - Input: Bosphorus 1080pabc150300450600750SE +/- 6.99, N = 15SE +/- 7.24, N = 3SE +/- 5.00, N = 15698.57699.65698.011. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
OpenBenchmarking.orgFrames Per Second, More Is BetterSVT-AV1 1.5Encoder Mode: Preset 13 - Input: Bosphorus 1080pabc120240360480600Min: 656.6 / Avg: 698.57 / Max: 760.6Min: 690.14 / Avg: 699.65 / Max: 713.86Min: 668.48 / Avg: 698.01 / Max: 742.371. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq

Faiss

Faiss is developed by Meta/Facebook. Faiss is a library for efficient similarity search and clustering of dense vectors. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgSeconds, Fewer Is BetterFaiss 1.7.4Test: demo_sift1Mabc1428425670SE +/- 0.03, N = 3SE +/- 0.02, N = 3SE +/- 0.13, N = 364.0163.9064.061. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgSeconds, Fewer Is BetterFaiss 1.7.4Test: demo_sift1Mabc1326395265Min: 63.94 / Avg: 64.01 / Max: 64.05Min: 63.88 / Avg: 63.9 / Max: 63.94Min: 63.89 / Avg: 64.06 / Max: 64.311. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc

OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - PQ baselineabc0.59111.18221.77332.36442.9555SE +/- 0.005, N = 3SE +/- 0.004, N = 3SE +/- 0.015, N = 32.6242.6132.6271. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 64abc0.92861.85722.78583.71444.643SE +/- 0.002, N = 3SE +/- 0.008, N = 3SE +/- 0.024, N = 34.1274.0914.1171. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 62abc0.78031.56062.34093.12123.9015SE +/- 0.002, N = 3SE +/- 0.002, N = 3SE +/- 0.021, N = 33.4663.4493.4681. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 58abc0.48020.96041.44061.92082.401SE +/- 0.001, N = 3SE +/- 0.006, N = 3SE +/- 0.013, N = 32.1322.1262.1341. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 54abc0.28910.57820.86731.15641.4455SE +/- 0.000, N = 3SE +/- 0.005, N = 3SE +/- 0.008, N = 31.2841.2781.2851. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 50abc0.1870.3740.5610.7480.935SE +/- 0.001, N = 3SE +/- 0.002, N = 3SE +/- 0.005, N = 30.8300.8250.8311. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 46abc0.13770.27540.41310.55080.6885SE +/- 0.000, N = 3SE +/- 0.003, N = 3SE +/- 0.004, N = 30.6120.6100.6121. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 42abc0.11810.23620.35430.47240.5905SE +/- 0.001, N = 3SE +/- 0.003, N = 3SE +/- 0.004, N = 30.5250.5220.5251. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 38abc0.11160.22320.33480.44640.558SE +/- 0.000, N = 3SE +/- 0.003, N = 3SE +/- 0.003, N = 30.4950.4940.4961. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 34abc0.10940.21880.32820.43760.547SE +/- 0.000, N = 3SE +/- 0.003, N = 3SE +/- 0.003, N = 30.4850.4850.4861. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - Polysemous 30abc0.10870.21740.32610.43480.5435SE +/- 0.001, N = 3SE +/- 0.003, N = 3SE +/- 0.003, N = 30.4820.4810.4831. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
OpenBenchmarking.orgms per query, Fewer Is BetterFaiss 1.7.4Test: bench_polysemous_sift1m - PQ baselineabc246810Min: 2.62 / Avg: 2.62 / Max: 2.63Min: 2.61 / Avg: 2.61 / Max: 2.62Min: 2.61 / Avg: 2.63 / Max: 2.661. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc

Intel TensorFlow

Intel optimized version of TensorFlow with benchmarks of Intel AI models and configurable batch sizes. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 1abc306090120150SE +/- 0.67, N = 3SE +/- 0.56, N = 3SE +/- 1.27, N = 3139.05139.52135.89
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 1abc306090120150Min: 137.75 / Avg: 139.05 / Max: 139.96Min: 138.47 / Avg: 139.52 / Max: 140.4Min: 134.34 / Avg: 135.89 / Max: 138.41

OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 1abc246810SE +/- 0.034, N = 3SE +/- 0.029, N = 3SE +/- 0.068, N = 37.1927.1687.360
OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 1abc3691215Min: 7.15 / Avg: 7.19 / Max: 7.26Min: 7.12 / Avg: 7.17 / Max: 7.22Min: 7.23 / Avg: 7.36 / Max: 7.44

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 1abc120240360480600SE +/- 5.71, N = 3SE +/- 5.31, N = 3SE +/- 11.17, N = 15547.78503.50523.17
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 1abc100200300400500Min: 538.31 / Avg: 547.78 / Max: 558.04Min: 493.05 / Avg: 503.5 / Max: 510.37Min: 448.35 / Avg: 523.17 / Max: 588.36

OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 1abc0.44690.89381.34071.78762.2345SE +/- 0.019, N = 3SE +/- 0.021, N = 3SE +/- 0.041, N = 151.8261.9861.924
OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 1abc246810Min: 1.79 / Avg: 1.83 / Max: 1.86Min: 1.96 / Avg: 1.99 / Max: 2.03Min: 1.7 / Avg: 1.92 / Max: 2.23

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 16abc60120180240300SE +/- 3.82, N = 3SE +/- 1.19, N = 3SE +/- 0.86, N = 3264.85267.53267.66
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 16abc50100150200250Min: 257.55 / Avg: 264.85 / Max: 270.43Min: 265.38 / Avg: 267.53 / Max: 269.49Min: 266.03 / Avg: 267.66 / Max: 268.95

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 32abc60120180240300SE +/- 0.79, N = 3SE +/- 0.40, N = 3SE +/- 0.53, N = 3252.89253.60252.68
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 32abc50100150200250Min: 251.32 / Avg: 252.89 / Max: 253.69Min: 252.82 / Avg: 253.6 / Max: 254.16Min: 251.65 / Avg: 252.68 / Max: 253.38

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 64abc50100150200250SE +/- 0.80, N = 3SE +/- 0.54, N = 3SE +/- 0.58, N = 3247.81247.96248.70
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 64abc50100150200250Min: 246.24 / Avg: 247.81 / Max: 248.83Min: 247.24 / Avg: 247.96 / Max: 249.03Min: 247.7 / Avg: 248.7 / Max: 249.71

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 96abc50100150200250SE +/- 0.71, N = 3SE +/- 1.07, N = 3SE +/- 0.56, N = 3245.03246.19245.61
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 96abc4080120160200Min: 243.84 / Avg: 245.03 / Max: 246.29Min: 244.14 / Avg: 246.19 / Max: 247.75Min: 244.71 / Avg: 245.61 / Max: 246.64

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 16abc2004006008001000SE +/- 3.01, N = 3SE +/- 2.79, N = 3SE +/- 4.43, N = 3800.10798.92793.97
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 16abc140280420560700Min: 795.41 / Avg: 800.1 / Max: 805.72Min: 793.48 / Avg: 798.92 / Max: 802.75Min: 785.78 / Avg: 793.97 / Max: 800.98

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 32abc2004006008001000SE +/- 2.87, N = 3SE +/- 1.36, N = 3SE +/- 2.90, N = 3824.03825.34830.17
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 32abc150300450600750Min: 819.63 / Avg: 824.03 / Max: 829.42Min: 822.61 / Avg: 825.34 / Max: 826.76Min: 824.99 / Avg: 830.17 / Max: 835.02

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 64abc2004006008001000SE +/- 1.51, N = 3SE +/- 0.81, N = 3SE +/- 2.49, N = 3830.21837.08833.73
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 64abc150300450600750Min: 827.22 / Avg: 830.2 / Max: 832.08Min: 836.2 / Avg: 837.08 / Max: 838.71Min: 830.07 / Avg: 833.73 / Max: 838.49

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 96abc2004006008001000SE +/- 1.57, N = 3SE +/- 1.99, N = 3SE +/- 0.36, N = 3803.35803.69805.52
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 96abc140280420560700Min: 800.32 / Avg: 803.35 / Max: 805.59Min: 799.71 / Avg: 803.69 / Max: 805.83Min: 804.85 / Avg: 805.52 / Max: 806.08

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 256abc50100150200250SE +/- 0.19, N = 3SE +/- 0.16, N = 3SE +/- 0.23, N = 3247.46247.09247.45
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 256abc4080120160200Min: 247.12 / Avg: 247.46 / Max: 247.76Min: 246.77 / Avg: 247.09 / Max: 247.25Min: 247 / Avg: 247.45 / Max: 247.75

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 512abc50100150200250SE +/- 0.07, N = 3SE +/- 0.12, N = 3SE +/- 0.10, N = 3249.62249.92249.76
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_fp32_pretrained_model - Batch Size: 512abc50100150200250Min: 249.48 / Avg: 249.62 / Max: 249.73Min: 249.68 / Avg: 249.92 / Max: 250.08Min: 249.67 / Avg: 249.76 / Max: 249.96

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 256abc2004006008001000SE +/- 1.40, N = 3SE +/- 2.42, N = 3SE +/- 0.79, N = 3779.33779.24780.45
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 256abc140280420560700Min: 776.73 / Avg: 779.33 / Max: 781.52Min: 774.75 / Avg: 779.24 / Max: 783.04Min: 778.9 / Avg: 780.45 / Max: 781.47

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 512abc2004006008001000SE +/- 0.87, N = 3SE +/- 0.47, N = 3SE +/- 0.23, N = 3774.18775.26775.56
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: resnet50_int8_pretrained_model - Batch Size: 512abc140280420560700Min: 772.56 / Avg: 774.18 / Max: 775.54Min: 774.62 / Avg: 775.26 / Max: 776.19Min: 775.1 / Avg: 775.56 / Max: 775.84

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 1abc1428425670SE +/- 0.31, N = 3SE +/- 0.59, N = 3SE +/- 1.97, N = 1561.6561.0057.42
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 1abc1224364860Min: 61.1 / Avg: 61.65 / Max: 62.17Min: 60.13 / Avg: 61 / Max: 62.13Min: 42.24 / Avg: 57.42 / Max: 62.59

OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 1abc48121620SE +/- 0.04, N = 3SE +/- 0.25, N = 3SE +/- 0.03, N = 1517.0717.3017.20
OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 1abc48121620Min: 16.99 / Avg: 17.07 / Max: 17.13Min: 16.96 / Avg: 17.3 / Max: 17.79Min: 16.98 / Avg: 17.2 / Max: 17.34

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 1abc20406080100SE +/- 0.54, N = 3SE +/- 0.36, N = 3SE +/- 3.79, N = 12104.96105.5099.00
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 1abc20406080100Min: 103.87 / Avg: 104.96 / Max: 105.57Min: 104.9 / Avg: 105.5 / Max: 106.16Min: 59.37 / Avg: 99 / Max: 106.22

OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 1abc3691215SE +/- 0.06, N = 3SE +/- 0.05, N = 3SE +/- 0.06, N = 1210.5610.3510.45
OpenBenchmarking.orgms, Fewer Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 1abc3691215Min: 10.45 / Avg: 10.56 / Max: 10.63Min: 10.28 / Avg: 10.35 / Max: 10.45Min: 10.05 / Avg: 10.44 / Max: 10.82

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 1abc30060090012001500SE +/- 2.41, N = 3SE +/- 3.66, N = 3SE +/- 0.80, N = 31591.731573.541606.41
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 1abc30060090012001500Min: 1588.14 / Avg: 1591.73 / Max: 1596.31Min: 1568.99 / Avg: 1573.54 / Max: 1580.78Min: 1604.96 / Avg: 1606.41 / Max: 1607.74

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 1abc10002000300040005000SE +/- 2.14, N = 3SE +/- 2.97, N = 3SE +/- 1.54, N = 34525.904539.904528.32
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 1abc8001600240032004000Min: 4522.01 / Avg: 4525.9 / Max: 4529.41Min: 4534.4 / Avg: 4539.9 / Max: 4544.61Min: 4525.25 / Avg: 4528.32 / Max: 4530.03

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 16abc20406080100SE +/- 0.15, N = 3SE +/- 0.74, N = 3SE +/- 0.63, N = 383.6383.2582.65
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 16abc1632486480Min: 83.39 / Avg: 83.63 / Max: 83.91Min: 81.97 / Avg: 83.25 / Max: 84.55Min: 81.41 / Avg: 82.65 / Max: 83.49

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 32abc20406080100SE +/- 0.10, N = 3SE +/- 0.26, N = 3SE +/- 0.41, N = 380.4280.5680.28
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 32abc1530456075Min: 80.31 / Avg: 80.42 / Max: 80.62Min: 80.17 / Avg: 80.56 / Max: 81.04Min: 79.52 / Avg: 80.28 / Max: 80.92

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 64abc20406080100SE +/- 0.16, N = 3SE +/- 0.10, N = 3SE +/- 0.18, N = 379.1579.5079.00
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 64abc1530456075Min: 78.84 / Avg: 79.15 / Max: 79.4Min: 79.31 / Avg: 79.5 / Max: 79.66Min: 78.73 / Avg: 79 / Max: 79.35

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 96abc20406080100SE +/- 0.14, N = 3SE +/- 0.09, N = 3SE +/- 0.16, N = 379.0579.4478.91
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 96abc1530456075Min: 78.78 / Avg: 79.05 / Max: 79.23Min: 79.33 / Avg: 79.44 / Max: 79.63Min: 78.72 / Avg: 78.91 / Max: 79.22

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 16abc70140210280350SE +/- 5.23, N = 15SE +/- 4.04, N = 15SE +/- 5.26, N = 12311.78319.25312.28
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 16abc60120180240300Min: 275.24 / Avg: 311.78 / Max: 335.67Min: 292.76 / Avg: 319.25 / Max: 336.02Min: 288.86 / Avg: 312.28 / Max: 334.76

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 32abc70140210280350SE +/- 3.42, N = 3SE +/- 0.69, N = 3SE +/- 3.93, N = 3312.25315.79311.68
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 32abc60120180240300Min: 305.56 / Avg: 312.25 / Max: 316.84Min: 315.1 / Avg: 315.79 / Max: 317.16Min: 304.05 / Avg: 311.68 / Max: 317.09

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 64abc70140210280350SE +/- 0.71, N = 3SE +/- 2.44, N = 3SE +/- 1.33, N = 3309.86309.26314.14
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 64abc60120180240300Min: 308.49 / Avg: 309.86 / Max: 310.85Min: 304.47 / Avg: 309.26 / Max: 312.45Min: 311.49 / Avg: 314.14 / Max: 315.6

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 96abc70140210280350SE +/- 2.02, N = 3SE +/- 1.90, N = 3SE +/- 1.86, N = 3300.91303.13297.39
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 96abc50100150200250Min: 297.88 / Avg: 300.91 / Max: 304.73Min: 299.35 / Avg: 303.13 / Max: 305.35Min: 293.81 / Avg: 297.39 / Max: 300.03

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 16abc2004006008001000SE +/- 0.39, N = 3SE +/- 0.41, N = 3SE +/- 0.62, N = 31107.681108.831107.45
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 16abc2004006008001000Min: 1107.08 / Avg: 1107.68 / Max: 1108.41Min: 1108.19 / Avg: 1108.83 / Max: 1109.6Min: 1106.76 / Avg: 1107.45 / Max: 1108.69

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 32abc2004006008001000SE +/- 0.85, N = 3SE +/- 1.16, N = 3SE +/- 0.84, N = 31047.321047.981048.20
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 32abc2004006008001000Min: 1045.76 / Avg: 1047.32 / Max: 1048.67Min: 1045.7 / Avg: 1047.98 / Max: 1049.49Min: 1046.72 / Avg: 1048.2 / Max: 1049.63

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 64abc2004006008001000SE +/- 0.93, N = 3SE +/- 0.28, N = 3SE +/- 0.12, N = 3952.20951.88951.85
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 64abc170340510680850Min: 950.73 / Avg: 952.2 / Max: 953.92Min: 951.5 / Avg: 951.88 / Max: 952.44Min: 951.65 / Avg: 951.85 / Max: 952.07

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 96abc2004006008001000SE +/- 0.59, N = 3SE +/- 0.57, N = 3SE +/- 0.63, N = 3914.91915.83915.29
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 96abc160320480640800Min: 913.77 / Avg: 914.91 / Max: 915.73Min: 914.87 / Avg: 915.83 / Max: 916.83Min: 914.08 / Avg: 915.29 / Max: 916.2

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 16abc10002000300040005000SE +/- 13.68, N = 3SE +/- 10.16, N = 3SE +/- 30.59, N = 34671.254689.944667.75
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 16abc8001600240032004000Min: 4654.8 / Avg: 4671.25 / Max: 4698.41Min: 4669.62 / Avg: 4689.94 / Max: 4700.38Min: 4606.74 / Avg: 4667.75 / Max: 4702.19

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 32abc11002200330044005500SE +/- 5.24, N = 3SE +/- 21.06, N = 3SE +/- 10.06, N = 35094.815078.875092.76
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 32abc9001800270036004500Min: 5086.02 / Avg: 5094.81 / Max: 5104.15Min: 5038.84 / Avg: 5078.87 / Max: 5110.24Min: 5072.71 / Avg: 5092.76 / Max: 5104.1

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 64abc9001800270036004500SE +/- 0.17, N = 3SE +/- 2.46, N = 3SE +/- 4.71, N = 34327.784330.684334.43
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 64abc8001600240032004000Min: 4327.54 / Avg: 4327.78 / Max: 4328.11Min: 4325.9 / Avg: 4330.68 / Max: 4334.11Min: 4327.94 / Avg: 4334.43 / Max: 4343.59

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 96abc9001800270036004500SE +/- 8.18, N = 3SE +/- 1.91, N = 3SE +/- 8.05, N = 34293.104285.764278.58
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 96abc7001400210028003500Min: 4277.02 / Avg: 4293.1 / Max: 4303.76Min: 4282.35 / Avg: 4285.76 / Max: 4288.94Min: 4263.05 / Avg: 4278.58 / Max: 4290.02

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 256abc20406080100SE +/- 0.15, N = 3SE +/- 0.09, N = 3SE +/- 0.04, N = 380.3680.4080.20
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 256abc1530456075Min: 80.14 / Avg: 80.36 / Max: 80.65Min: 80.24 / Avg: 80.4 / Max: 80.56Min: 80.15 / Avg: 80.2 / Max: 80.27

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 512abc20406080100SE +/- 0.06, N = 3SE +/- 0.11, N = 3SE +/- 0.06, N = 381.0481.0780.95
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_fp32_pretrained_model - Batch Size: 512abc1530456075Min: 80.93 / Avg: 81.04 / Max: 81.12Min: 80.89 / Avg: 81.07 / Max: 81.26Min: 80.86 / Avg: 80.95 / Max: 81.06

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 256abc60120180240300SE +/- 0.52, N = 3SE +/- 0.61, N = 3SE +/- 0.38, N = 3288.45289.43287.48
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 256abc50100150200250Min: 287.42 / Avg: 288.45 / Max: 289.06Min: 288.27 / Avg: 289.43 / Max: 290.33Min: 286.73 / Avg: 287.48 / Max: 287.94

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 512abc60120180240300SE +/- 0.57, N = 3SE +/- 0.10, N = 3SE +/- 0.94, N = 3290.66288.25288.82
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: inceptionv4_int8_pretrained_model - Batch Size: 512abc50100150200250Min: 289.55 / Avg: 290.66 / Max: 291.47Min: 288.06 / Avg: 288.25 / Max: 288.39Min: 287.01 / Avg: 288.82 / Max: 290.15

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 256abc2004006008001000SE +/- 0.47, N = 3SE +/- 0.10, N = 3SE +/- 1.00, N = 3867.02867.13866.92
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 256abc150300450600750Min: 866.2 / Avg: 867.02 / Max: 867.84Min: 866.96 / Avg: 867.13 / Max: 867.29Min: 864.96 / Avg: 866.92 / Max: 868.2

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 512ac2004006008001000SE +/- 0.07, N = 3SE +/- 0.15, N = 3856.50855.61
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_fp32_pretrained_model - Batch Size: 512ac150300450600750Min: 856.37 / Avg: 856.5 / Max: 856.57Min: 855.31 / Avg: 855.61 / Max: 855.8

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 256ac7001400210028003500SE +/- 2.91, N = 3SE +/- 3.88, N = 33482.013482.91
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 256ac6001200180024003000Min: 3476.28 / Avg: 3482.01 / Max: 3485.8Min: 3476.52 / Avg: 3482.91 / Max: 3489.91

OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 512ac7001400210028003500SE +/- 1.83, N = 3SE +/- 1.99, N = 33263.943272.66
OpenBenchmarking.orgimages/sec, More Is BetterIntel TensorFlow 2.12Model: mobilenetv1_int8_pretrained_model - Batch Size: 512ac6001200180024003000Min: 3260.28 / Avg: 3263.94 / Max: 3265.93Min: 3269.69 / Avg: 3272.66 / Max: 3276.44

AMD ZenDNN TensorFlow

AMD ZenDNN optimized version of TensorFlow with various AMD UIF models. For this test profile to work you must first download TF_v2.10_ZenDNN_v4.0_Python_v3.10.zip from https://www.amd.com/en/developer/zendnn.html and put it in your Phoronix Test Suite download cache. Learn more via the OpenBenchmarking.org test page.

OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 1ac246810SE +/- 0.015, N = 3SE +/- 0.051, N = 38.0708.122
OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 1ac3691215Min: 8.05 / Avg: 8.07 / Max: 8.1Min: 8.03 / Avg: 8.12 / Max: 8.2

OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 1ac612182430SE +/- 0.01, N = 3SE +/- 0.02, N = 323.1623.13
OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 1ac510152025Min: 23.14 / Avg: 23.16 / Max: 23.18Min: 23.11 / Avg: 23.13 / Max: 23.17

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 16ac50100150200250SE +/- 0.17, N = 3SE +/- 0.45, N = 3241.34240.72
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 16ac4080120160200Min: 241.08 / Avg: 241.34 / Max: 241.65Min: 239.87 / Avg: 240.72 / Max: 241.4

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 32ac50100150200250SE +/- 0.13, N = 3SE +/- 1.39, N = 3228.35226.86
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 32ac4080120160200Min: 228.19 / Avg: 228.35 / Max: 228.6Min: 224.08 / Avg: 226.86 / Max: 228.33

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 64ac50100150200250SE +/- 0.32, N = 3SE +/- 1.23, N = 3225.81226.03
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 64ac4080120160200Min: 225.39 / Avg: 225.81 / Max: 226.44Min: 223.86 / Avg: 226.03 / Max: 228.12

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 96ac50100150200250SE +/- 1.65, N = 3SE +/- 0.11, N = 3228.11230.14
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 96ac4080120160200Min: 224.81 / Avg: 228.11 / Max: 229.76Min: 229.92 / Avg: 230.14 / Max: 230.3

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 16ac1428425670SE +/- 0.01, N = 3SE +/- 0.03, N = 362.7862.79
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 16ac1224364860Min: 62.76 / Avg: 62.78 / Max: 62.8Min: 62.75 / Avg: 62.79 / Max: 62.85

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 32ac1428425670SE +/- 0.11, N = 3SE +/- 0.06, N = 362.8262.91
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 32ac1224364860Min: 62.67 / Avg: 62.82 / Max: 63.03Min: 62.83 / Avg: 62.91 / Max: 63.03

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 64ac1428425670SE +/- 0.12, N = 3SE +/- 0.07, N = 364.6364.61
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 64ac1326395265Min: 64.39 / Avg: 64.63 / Max: 64.8Min: 64.49 / Avg: 64.61 / Max: 64.73

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 96ac1530456075SE +/- 0.06, N = 3SE +/- 0.19, N = 365.8665.59
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 96ac1326395265Min: 65.77 / Avg: 65.86 / Max: 65.98Min: 65.2 / Avg: 65.59 / Max: 65.79

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 256ac50100150200250SE +/- 0.86, N = 3SE +/- 0.13, N = 3238.11239.23
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 256ac4080120160200Min: 236.74 / Avg: 238.11 / Max: 239.7Min: 239.08 / Avg: 239.23 / Max: 239.49

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 512ac50100150200250SE +/- 0.55, N = 3SE +/- 0.21, N = 3242.65243.29
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 512ac4080120160200Min: 241.57 / Avg: 242.65 / Max: 243.41Min: 242.87 / Avg: 243.29 / Max: 243.53

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 256ac1530456075SE +/- 0.29, N = 3SE +/- 0.04, N = 368.2468.42
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 256ac1326395265Min: 67.68 / Avg: 68.24 / Max: 68.65Min: 68.38 / Avg: 68.42 / Max: 68.5

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 512ac1632486480SE +/- 0.01, N = 3SE +/- 0.08, N = 370.3670.06
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 512ac1428425670Min: 70.34 / Avg: 70.36 / Max: 70.38Min: 69.94 / Avg: 70.06 / Max: 70.21

OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 1ac0.35690.71381.07071.42761.7845SE +/- 0.021, N = 15SE +/- 0.020, N = 151.5861.573
OpenBenchmarking.orgms, Fewer Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 1ac246810Min: 1.46 / Avg: 1.59 / Max: 1.66Min: 1.46 / Avg: 1.57 / Max: 1.64

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 16ac2004006008001000SE +/- 1.60, N = 3SE +/- 1.06, N = 3953.78953.58
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 16ac2004006008001000Min: 951.54 / Avg: 953.78 / Max: 956.88Min: 951.52 / Avg: 953.58 / Max: 955.02

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 32ac170340510680850SE +/- 0.44, N = 3SE +/- 0.39, N = 3770.81768.31
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 32ac140280420560700Min: 769.93 / Avg: 770.81 / Max: 771.34Min: 767.64 / Avg: 768.31 / Max: 768.99

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 64ac150300450600750SE +/- 2.05, N = 3SE +/- 1.36, N = 3715.78717.29
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 64ac130260390520650Min: 711.77 / Avg: 715.78 / Max: 718.5Min: 714.58 / Avg: 717.29 / Max: 718.78

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 96ac150300450600750SE +/- 0.52, N = 3SE +/- 0.56, N = 3691.01690.86
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 96ac120240360480600Min: 689.99 / Avg: 691.01 / Max: 691.71Min: 690.23 / Avg: 690.86 / Max: 691.98

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 256ac140280420560700SE +/- 0.43, N = 3SE +/- 0.46, N = 3650.17650.04
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 256ac110220330440550Min: 649.5 / Avg: 650.17 / Max: 650.96Min: 649.15 / Avg: 650.04 / Max: 650.7

OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 512ac140280420560700SE +/- 0.23, N = 3SE +/- 0.86, N = 3643.44642.22
OpenBenchmarking.orgimages/sec, More Is BetterAMD ZenDNN TensorFlow 2.10 ZenDNN 4.0Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 512ac110220330440550Min: 643.18 / Avg: 643.44 / Max: 643.9Min: 640.99 / Avg: 642.22 / Max: 643.88

77 Results Shown

SVT-AV1:
  Preset 4 - Bosphorus 4K
  Preset 8 - Bosphorus 4K
  Preset 12 - Bosphorus 4K
  Preset 13 - Bosphorus 4K
  Preset 4 - Bosphorus 1080p
  Preset 8 - Bosphorus 1080p
  Preset 12 - Bosphorus 1080p
  Preset 13 - Bosphorus 1080p
Faiss:
  demo_sift1M
  bench_polysemous_sift1m - PQ baseline
Intel TensorFlow:
  resnet50_fp32_pretrained_model - 1:
    images/sec
    ms
  resnet50_int8_pretrained_model - 1:
    images/sec
    ms
  resnet50_fp32_pretrained_model - 16:
    images/sec
  resnet50_fp32_pretrained_model - 32:
    images/sec
  resnet50_fp32_pretrained_model - 64:
    images/sec
  resnet50_fp32_pretrained_model - 96:
    images/sec
  resnet50_int8_pretrained_model - 16:
    images/sec
  resnet50_int8_pretrained_model - 32:
    images/sec
  resnet50_int8_pretrained_model - 64:
    images/sec
  resnet50_int8_pretrained_model - 96:
    images/sec
  resnet50_fp32_pretrained_model - 256:
    images/sec
  resnet50_fp32_pretrained_model - 512:
    images/sec
  resnet50_int8_pretrained_model - 256:
    images/sec
  resnet50_int8_pretrained_model - 512:
    images/sec
  inceptionv4_fp32_pretrained_model - 1:
    images/sec
    ms
  inceptionv4_int8_pretrained_model - 1:
    images/sec
    ms
  mobilenetv1_fp32_pretrained_model - 1:
    images/sec
  mobilenetv1_int8_pretrained_model - 1:
    images/sec
  inceptionv4_fp32_pretrained_model - 16:
    images/sec
  inceptionv4_fp32_pretrained_model - 32:
    images/sec
  inceptionv4_fp32_pretrained_model - 64:
    images/sec
  inceptionv4_fp32_pretrained_model - 96:
    images/sec
  inceptionv4_int8_pretrained_model - 16:
    images/sec
  inceptionv4_int8_pretrained_model - 32:
    images/sec
  inceptionv4_int8_pretrained_model - 64:
    images/sec
  inceptionv4_int8_pretrained_model - 96:
    images/sec
  mobilenetv1_fp32_pretrained_model - 16:
    images/sec
  mobilenetv1_fp32_pretrained_model - 32:
    images/sec
  mobilenetv1_fp32_pretrained_model - 64:
    images/sec
  mobilenetv1_fp32_pretrained_model - 96:
    images/sec
  mobilenetv1_int8_pretrained_model - 16:
    images/sec
  mobilenetv1_int8_pretrained_model - 32:
    images/sec
  mobilenetv1_int8_pretrained_model - 64:
    images/sec
  mobilenetv1_int8_pretrained_model - 96:
    images/sec
  inceptionv4_fp32_pretrained_model - 256:
    images/sec
  inceptionv4_fp32_pretrained_model - 512:
    images/sec
  inceptionv4_int8_pretrained_model - 256:
    images/sec
  inceptionv4_int8_pretrained_model - 512:
    images/sec
  mobilenetv1_fp32_pretrained_model - 256:
    images/sec
  mobilenetv1_fp32_pretrained_model - 512:
    images/sec
  mobilenetv1_int8_pretrained_model - 256:
    images/sec
  mobilenetv1_int8_pretrained_model - 512:
    images/sec
AMD ZenDNN TensorFlow:
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 1
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 1
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 16
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 32
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 64
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 96
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 16
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 32
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 64
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 96
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 256
  tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 512
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 256
  tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 512
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 1
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 16
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 32
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 64
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 96
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 256
  tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 512

a

Kernel Notes: Transparent Huge Pages: madvise
Compiler 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-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 Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203
Python Notes: Python 3.10.9
Security Notes: 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 + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 28 April 2023 10:05 by user pts.

b

Kernel Notes: Transparent Huge Pages: madvise
Compiler 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-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 Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203
Python Notes: Python 3.10.9
Security Notes: 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 + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 28 April 2023 19:21 by user pts.

c

Processor: AMD Ryzen 9 7950X 16-Core @ 4.50GHz (16 Cores / 32 Threads), Motherboard: ASUS ROG CROSSHAIR X670E HERO (1101 BIOS), Chipset: AMD Device 14d8, Memory: 32GB, Disk: 2048GB SOLIDIGM SSDPFKKW020X7 + 2000GB, Graphics: AMD Radeon RX 7900 XTX 24GB (2304/1249MHz), Audio: AMD Device ab30, Monitor: ASUS MG28U, Network: Intel I225-V + Intel Wi-Fi 6 AX210/AX211/AX411

OS: Ubuntu 22.04, Kernel: 6.3.0-060300rc7daily20230417-generic (x86_64), Desktop: GNOME Shell 42.5, Display Server: X Server 1.21.1.3 + Wayland, OpenGL: 4.6 Mesa 23.2.0-devel (git-f6fb189 2023-04-18 jammy-oibaf-ppa) (LLVM 15.0.7 DRM 3.52), Vulkan: 1.3.246, Compiler: GCC 11.3.0, File-System: ext4, Screen Resolution: 3840x2160

Kernel Notes: Transparent Huge Pages: madvise
Compiler 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-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 Notes: Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203
Python Notes: Python 3.10.9
Security Notes: 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 + spectre_v1: Mitigation of usercopy/swapgs barriers and __user pointer sanitization + spectre_v2: Mitigation of Enhanced / Automatic IBRS IBPB: conditional RSB filling PBRSB-eIBRS: Not affected + srbds: Not affected + tsx_async_abort: Not affected

Testing initiated at 28 April 2023 21:51 by user pts.