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.
HTML result view exported from: https://openbenchmarking.org/result/2304292-PTS-NEWAI12536&rdt&grs .
new ai Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server OpenGL Vulkan Compiler File-System Screen Resolution a b c AMD Ryzen 9 7950X 16-Core @ 4.50GHz (16 Cores / 32 Threads) ASUS ROG CROSSHAIR X670E HERO (1101 BIOS) AMD Device 14d8 32GB 2048GB SOLIDIGM SSDPFKKW020X7 + 2000GB AMD Radeon RX 7900 XTX 24GB (2304/1249MHz) AMD Device ab30 ASUS MG28U Intel I225-V + Intel Wi-Fi 6 AX210/AX211/AX411 Ubuntu 22.04 6.3.0-060300rc7daily20230417-generic (x86_64) GNOME Shell 42.5 X Server 1.21.1.3 + Wayland 4.6 Mesa 23.2.0-devel (git-f6fb189 2023-04-18 jammy-oibaf-ppa) (LLVM 15.0.7 DRM 3.52) 1.3.246 GCC 11.3.0 ext4 3840x2160 OpenBenchmarking.org Kernel Details - Transparent Huge Pages: madvise Compiler Details - --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-bootstrap --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,brig,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-link-serialization=2 --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-targets=nvptx-none=/build/gcc-11-xKiWfi/gcc-11-11.3.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-11-xKiWfi/gcc-11-11.3.0/debian/tmp-gcn/usr --enable-plugin --enable-shared --enable-threads=posix --host=x86_64-linux-gnu --program-prefix=x86_64-linux-gnu- --target=x86_64-linux-gnu --with-abi=m64 --with-arch-32=i686 --with-build-config=bootstrap-lto-lean --with-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v Processor Details - Scaling Governor: acpi-cpufreq schedutil (Boost: Enabled) - CPU Microcode: 0xa601203 Python Details - Python 3.10.9 Security Details - itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + retbleed: Not affected + spec_store_bypass: Mitigation of SSB disabled via prctl + 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
new ai intel-tensorflow: resnet50_fp32_pretrained_model - 1 intel-tensorflow: resnet50_fp32_pretrained_model - 1 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 1 intel-tensorflow: inceptionv4_int8_pretrained_model - 1 svt-av1: Preset 8 - Bosphorus 1080p intel-tensorflow: inceptionv4_int8_pretrained_model - 96 intel-tensorflow: inceptionv4_int8_pretrained_model - 64 intel-tensorflow: inceptionv4_fp32_pretrained_model - 1 intel-tensorflow: inceptionv4_int8_pretrained_model - 32 svt-av1: Preset 12 - Bosphorus 4K intel-tensorflow: inceptionv4_fp32_pretrained_model - 16 intel-tensorflow: resnet50_fp32_pretrained_model - 16 svt-av1: Preset 4 - Bosphorus 1080p zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 96 svt-av1: Preset 4 - Bosphorus 4K intel-tensorflow: inceptionv4_int8_pretrained_model - 512 intel-tensorflow: resnet50_int8_pretrained_model - 64 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 1 svt-av1: Preset 12 - Bosphorus 1080p intel-tensorflow: resnet50_int8_pretrained_model - 16 intel-tensorflow: resnet50_int8_pretrained_model - 32 intel-tensorflow: inceptionv4_int8_pretrained_model - 256 intel-tensorflow: inceptionv4_fp32_pretrained_model - 96 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 32 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 1 intel-tensorflow: inceptionv4_fp32_pretrained_model - 64 faiss: bench_polysemous_sift1m - PQ baseline intel-tensorflow: mobilenetv1_int8_pretrained_model - 16 intel-tensorflow: resnet50_fp32_pretrained_model - 96 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 256 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 512 svt-av1: Preset 13 - Bosphorus 4K zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 96 svt-av1: Preset 8 - Bosphorus 4K intel-tensorflow: resnet50_fp32_pretrained_model - 32 intel-tensorflow: resnet50_fp32_pretrained_model - 64 intel-tensorflow: inceptionv4_fp32_pretrained_model - 32 intel-tensorflow: mobilenetv1_int8_pretrained_model - 96 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 32 intel-tensorflow: mobilenetv1_int8_pretrained_model - 32 intel-tensorflow: mobilenetv1_int8_pretrained_model - 1 intel-tensorflow: resnet50_int8_pretrained_model - 96 intel-tensorflow: mobilenetv1_int8_pretrained_model - 512 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 256 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 512 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 16 intel-tensorflow: inceptionv4_fp32_pretrained_model - 256 faiss: demo_sift1M svt-av1: Preset 13 - Bosphorus 1080p zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 64 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 512 intel-tensorflow: resnet50_int8_pretrained_model - 512 intel-tensorflow: resnet50_int8_pretrained_model - 256 intel-tensorflow: mobilenetv1_int8_pretrained_model - 64 intel-tensorflow: resnet50_fp32_pretrained_model - 256 intel-tensorflow: inceptionv4_fp32_pretrained_model - 512 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 32 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 16 intel-tensorflow: resnet50_fp32_pretrained_model - 512 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 1 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 512 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 96 zendnn-tensorflow: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - 64 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 32 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 64 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 64 intel-tensorflow: mobilenetv1_int8_pretrained_model - 256 intel-tensorflow: mobilenetv1_fp32_pretrained_model - 256 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 96 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 16 zendnn-tensorflow: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - 256 zendnn-tensorflow: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - 16 intel-tensorflow: inceptionv4_int8_pretrained_model - 16 intel-tensorflow: inceptionv4_int8_pretrained_model - 1 intel-tensorflow: inceptionv4_fp32_pretrained_model - 1 intel-tensorflow: resnet50_int8_pretrained_model - 1 intel-tensorflow: resnet50_int8_pretrained_model - 1 faiss: bench_polysemous_sift1m - Polysemous 30 faiss: bench_polysemous_sift1m - Polysemous 34 faiss: bench_polysemous_sift1m - Polysemous 38 faiss: bench_polysemous_sift1m - Polysemous 42 faiss: bench_polysemous_sift1m - Polysemous 46 faiss: bench_polysemous_sift1m - Polysemous 50 faiss: bench_polysemous_sift1m - Polysemous 54 faiss: bench_polysemous_sift1m - Polysemous 58 faiss: bench_polysemous_sift1m - Polysemous 62 faiss: bench_polysemous_sift1m - Polysemous 64 a b c 7.192 139.051 1591.73 10.560 119.996 300.91 309.86 17.070 312.25 210.908 83.63 264.850 14.168 228.11 6.111 290.66 830.205 1.586 716.460 800.103 824.027 288.45 79.05 228.35 8.070 79.15 2.624 4671.25 245.027 238.11 70.36 206.520 65.86 75.893 252.894 247.813 80.42 4293.10 770.81 5094.81 4525.90 803.346 3263.94 68.24 242.65 241.34 80.36 64.009 698.567 715.78 643.44 774.177 779.333 4327.78 247.463 81.04 62.82 1107.68 249.615 23.159 856.50 914.91 225.81 1047.32 952.20 64.63 3482.01 867.02 691.01 953.78 650.17 62.78 311.78 104.96 61.65 1.826 547.777 0.482 0.485 0.495 0.525 0.612 0.830 1.284 2.132 3.466 4.127 7.168 139.519 1573.54 10.351 120.306 303.13 309.26 17.296 315.79 213.399 83.25 267.531 14.091 6.165 288.25 837.083 711.29 798.920 825.337 289.43 79.44 79.50 2.613 4689.94 246.187 207.373 75.807 253.600 247.959 80.56 4285.76 5078.87 4539.90 803.687 80.40 63.903 699.652 775.258 779.241 4330.68 247.089 81.07 1108.83 249.918 915.83 1047.98 951.88 867.13 319.25 105.50 61.00 1.986 503.500 0.481 0.485 0.494 0.522 0.610 0.825 1.278 2.126 3.449 4.091 7.360 135.894 1606.41 10.445 118.000 297.39 314.14 17.200 311.68 210.656 82.65 267.660 14.234 230.14 6.111 288.82 833.729 1.573 710.846 793.970 830.170 287.48 78.91 226.86 8.122 79.00 2.627 4667.75 245.605 239.23 70.06 206.943 65.59 75.612 252.684 248.704 80.28 4278.58 768.31 5092.76 4528.32 805.521 3272.66 68.42 243.29 240.72 80.20 64.060 698.009 717.29 642.22 775.557 780.454 4334.43 247.448 80.95 62.91 1107.45 249.763 23.131 855.61 915.29 226.03 1048.20 951.85 64.61 3482.91 866.92 690.86 953.58 650.04 62.79 312.28 99.00 57.42 1.924 523.171 0.483 0.486 0.496 0.525 0.612 0.831 1.285 2.134 3.468 4.117 OpenBenchmarking.org
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 1 a b c 2 4 6 8 10 SE +/- 0.034, N = 3 SE +/- 0.029, N = 3 SE +/- 0.068, N = 3 7.192 7.168 7.360
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 1 a b c 30 60 90 120 150 SE +/- 0.67, N = 3 SE +/- 0.56, N = 3 SE +/- 1.27, N = 3 139.05 139.52 135.89
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 1 a b c 300 600 900 1200 1500 SE +/- 2.41, N = 3 SE +/- 3.66, N = 3 SE +/- 0.80, N = 3 1591.73 1573.54 1606.41
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 1 a b c 3 6 9 12 15 SE +/- 0.06, N = 3 SE +/- 0.05, N = 3 SE +/- 0.06, N = 12 10.56 10.35 10.45
SVT-AV1 Encoder Mode: Preset 8 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 8 - Input: Bosphorus 1080p a b c 30 60 90 120 150 SE +/- 0.60, N = 3 SE +/- 0.75, N = 3 SE +/- 0.63, N = 3 120.00 120.31 118.00 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 96 a b c 70 140 210 280 350 SE +/- 2.02, N = 3 SE +/- 1.90, N = 3 SE +/- 1.86, N = 3 300.91 303.13 297.39
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 64 a b c 70 140 210 280 350 SE +/- 0.71, N = 3 SE +/- 2.44, N = 3 SE +/- 1.33, N = 3 309.86 309.26 314.14
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 1 a b c 4 8 12 16 20 SE +/- 0.04, N = 3 SE +/- 0.25, N = 3 SE +/- 0.03, N = 15 17.07 17.30 17.20
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 32 a b c 70 140 210 280 350 SE +/- 3.42, N = 3 SE +/- 0.69, N = 3 SE +/- 3.93, N = 3 312.25 315.79 311.68
SVT-AV1 Encoder Mode: Preset 12 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 12 - Input: Bosphorus 4K a b c 50 100 150 200 250 SE +/- 1.35, N = 3 SE +/- 0.58, N = 3 SE +/- 0.55, N = 3 210.91 213.40 210.66 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 16 a b c 20 40 60 80 100 SE +/- 0.15, N = 3 SE +/- 0.74, N = 3 SE +/- 0.63, N = 3 83.63 83.25 82.65
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 16 a b c 60 120 180 240 300 SE +/- 3.82, N = 3 SE +/- 1.19, N = 3 SE +/- 0.86, N = 3 264.85 267.53 267.66
SVT-AV1 Encoder Mode: Preset 4 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 4 - Input: Bosphorus 1080p a b c 4 8 12 16 20 SE +/- 0.02, N = 3 SE +/- 0.08, N = 3 SE +/- 0.05, N = 3 14.17 14.09 14.23 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 96 a c 50 100 150 200 250 SE +/- 1.65, N = 3 SE +/- 0.11, N = 3 228.11 230.14
SVT-AV1 Encoder Mode: Preset 4 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 4 - Input: Bosphorus 4K a b c 2 4 6 8 10 SE +/- 0.030, N = 3 SE +/- 0.017, N = 3 SE +/- 0.004, N = 3 6.111 6.165 6.111 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 512 a b c 60 120 180 240 300 SE +/- 0.57, N = 3 SE +/- 0.10, N = 3 SE +/- 0.94, N = 3 290.66 288.25 288.82
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 64 a b c 200 400 600 800 1000 SE +/- 1.51, N = 3 SE +/- 0.81, N = 3 SE +/- 2.49, N = 3 830.21 837.08 833.73
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 1 a c 0.3569 0.7138 1.0707 1.4276 1.7845 SE +/- 0.021, N = 15 SE +/- 0.020, N = 15 1.586 1.573
SVT-AV1 Encoder Mode: Preset 12 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 12 - Input: Bosphorus 1080p a b c 150 300 450 600 750 SE +/- 5.15, N = 3 SE +/- 8.08, N = 3 SE +/- 6.31, N = 3 716.46 711.29 710.85 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 16 a b c 200 400 600 800 1000 SE +/- 3.01, N = 3 SE +/- 2.79, N = 3 SE +/- 4.43, N = 3 800.10 798.92 793.97
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 32 a b c 200 400 600 800 1000 SE +/- 2.87, N = 3 SE +/- 1.36, N = 3 SE +/- 2.90, N = 3 824.03 825.34 830.17
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 256 a b c 60 120 180 240 300 SE +/- 0.52, N = 3 SE +/- 0.61, N = 3 SE +/- 0.38, N = 3 288.45 289.43 287.48
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 96 a b c 20 40 60 80 100 SE +/- 0.14, N = 3 SE +/- 0.09, N = 3 SE +/- 0.16, N = 3 79.05 79.44 78.91
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 32 a c 50 100 150 200 250 SE +/- 0.13, N = 3 SE +/- 1.39, N = 3 228.35 226.86
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 1 a c 2 4 6 8 10 SE +/- 0.015, N = 3 SE +/- 0.051, N = 3 8.070 8.122
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 64 a b c 20 40 60 80 100 SE +/- 0.16, N = 3 SE +/- 0.10, N = 3 SE +/- 0.18, N = 3 79.15 79.50 79.00
Faiss Test: bench_polysemous_sift1m - PQ baseline OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - PQ baseline a b c 0.5911 1.1822 1.7733 2.3644 2.9555 SE +/- 0.005, N = 3 SE +/- 0.004, N = 3 SE +/- 0.015, N = 3 2.624 2.613 2.627 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 16 a b c 1000 2000 3000 4000 5000 SE +/- 13.68, N = 3 SE +/- 10.16, N = 3 SE +/- 30.59, N = 3 4671.25 4689.94 4667.75
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 96 a b c 50 100 150 200 250 SE +/- 0.71, N = 3 SE +/- 1.07, N = 3 SE +/- 0.56, N = 3 245.03 246.19 245.61
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 256 a c 50 100 150 200 250 SE +/- 0.86, N = 3 SE +/- 0.13, N = 3 238.11 239.23
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 512 a c 16 32 48 64 80 SE +/- 0.01, N = 3 SE +/- 0.08, N = 3 70.36 70.06
SVT-AV1 Encoder Mode: Preset 13 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 13 - Input: Bosphorus 4K a b c 50 100 150 200 250 SE +/- 1.23, N = 3 SE +/- 0.95, N = 3 SE +/- 0.70, N = 3 206.52 207.37 206.94 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 96 a c 15 30 45 60 75 SE +/- 0.06, N = 3 SE +/- 0.19, N = 3 65.86 65.59
SVT-AV1 Encoder Mode: Preset 8 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 8 - Input: Bosphorus 4K a b c 20 40 60 80 100 SE +/- 0.39, N = 3 SE +/- 0.11, N = 3 SE +/- 0.56, N = 3 75.89 75.81 75.61 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 32 a b c 60 120 180 240 300 SE +/- 0.79, N = 3 SE +/- 0.40, N = 3 SE +/- 0.53, N = 3 252.89 253.60 252.68
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 64 a b c 50 100 150 200 250 SE +/- 0.80, N = 3 SE +/- 0.54, N = 3 SE +/- 0.58, N = 3 247.81 247.96 248.70
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 32 a b c 20 40 60 80 100 SE +/- 0.10, N = 3 SE +/- 0.26, N = 3 SE +/- 0.41, N = 3 80.42 80.56 80.28
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 96 a b c 900 1800 2700 3600 4500 SE +/- 8.18, N = 3 SE +/- 1.91, N = 3 SE +/- 8.05, N = 3 4293.10 4285.76 4278.58
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 32 a c 170 340 510 680 850 SE +/- 0.44, N = 3 SE +/- 0.39, N = 3 770.81 768.31
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 32 a b c 1100 2200 3300 4400 5500 SE +/- 5.24, N = 3 SE +/- 21.06, N = 3 SE +/- 10.06, N = 3 5094.81 5078.87 5092.76
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 1 a b c 1000 2000 3000 4000 5000 SE +/- 2.14, N = 3 SE +/- 2.97, N = 3 SE +/- 1.54, N = 3 4525.90 4539.90 4528.32
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 96 a b c 200 400 600 800 1000 SE +/- 1.57, N = 3 SE +/- 1.99, N = 3 SE +/- 0.36, N = 3 803.35 803.69 805.52
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 512 a c 700 1400 2100 2800 3500 SE +/- 1.83, N = 3 SE +/- 1.99, N = 3 3263.94 3272.66
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 256 a c 15 30 45 60 75 SE +/- 0.29, N = 3 SE +/- 0.04, N = 3 68.24 68.42
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 512 a c 50 100 150 200 250 SE +/- 0.55, N = 3 SE +/- 0.21, N = 3 242.65 243.29
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 16 a c 50 100 150 200 250 SE +/- 0.17, N = 3 SE +/- 0.45, N = 3 241.34 240.72
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 256 a b c 20 40 60 80 100 SE +/- 0.15, N = 3 SE +/- 0.09, N = 3 SE +/- 0.04, N = 3 80.36 80.40 80.20
Faiss Test: demo_sift1M OpenBenchmarking.org Seconds, Fewer Is Better Faiss 1.7.4 Test: demo_sift1M a b c 14 28 42 56 70 SE +/- 0.03, N = 3 SE +/- 0.02, N = 3 SE +/- 0.13, N = 3 64.01 63.90 64.06 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
SVT-AV1 Encoder Mode: Preset 13 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 1.5 Encoder Mode: Preset 13 - Input: Bosphorus 1080p a b c 150 300 450 600 750 SE +/- 6.99, N = 15 SE +/- 7.24, N = 3 SE +/- 5.00, N = 15 698.57 699.65 698.01 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 64 a c 150 300 450 600 750 SE +/- 2.05, N = 3 SE +/- 1.36, N = 3 715.78 717.29
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 512 a c 140 280 420 560 700 SE +/- 0.23, N = 3 SE +/- 0.86, N = 3 643.44 642.22
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 512 a b c 200 400 600 800 1000 SE +/- 0.87, N = 3 SE +/- 0.47, N = 3 SE +/- 0.23, N = 3 774.18 775.26 775.56
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 256 a b c 200 400 600 800 1000 SE +/- 1.40, N = 3 SE +/- 2.42, N = 3 SE +/- 0.79, N = 3 779.33 779.24 780.45
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 64 a b c 900 1800 2700 3600 4500 SE +/- 0.17, N = 3 SE +/- 2.46, N = 3 SE +/- 4.71, N = 3 4327.78 4330.68 4334.43
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 256 a b c 50 100 150 200 250 SE +/- 0.19, N = 3 SE +/- 0.16, N = 3 SE +/- 0.23, N = 3 247.46 247.09 247.45
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 512 a b c 20 40 60 80 100 SE +/- 0.06, N = 3 SE +/- 0.11, N = 3 SE +/- 0.06, N = 3 81.04 81.07 80.95
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 32 a c 14 28 42 56 70 SE +/- 0.11, N = 3 SE +/- 0.06, N = 3 62.82 62.91
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 16 a b c 200 400 600 800 1000 SE +/- 0.39, N = 3 SE +/- 0.41, N = 3 SE +/- 0.62, N = 3 1107.68 1108.83 1107.45
Intel TensorFlow Model: resnet50_fp32_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_fp32_pretrained_model - Batch Size: 512 a b c 50 100 150 200 250 SE +/- 0.07, N = 3 SE +/- 0.12, N = 3 SE +/- 0.10, N = 3 249.62 249.92 249.76
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 1 a c 6 12 18 24 30 SE +/- 0.01, N = 3 SE +/- 0.02, N = 3 23.16 23.13
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 512 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 512 a c 200 400 600 800 1000 SE +/- 0.07, N = 3 SE +/- 0.15, N = 3 856.50 855.61
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 96 a b c 200 400 600 800 1000 SE +/- 0.59, N = 3 SE +/- 0.57, N = 3 SE +/- 0.63, N = 3 914.91 915.83 915.29
AMD ZenDNN TensorFlow Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_resnetv1_50_imagenet_224_224_6.97G_1.1_Z4.0 - Batch Size: 64 a c 50 100 150 200 250 SE +/- 0.32, N = 3 SE +/- 1.23, N = 3 225.81 226.03
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 32 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 32 a b c 200 400 600 800 1000 SE +/- 0.85, N = 3 SE +/- 1.16, N = 3 SE +/- 0.84, N = 3 1047.32 1047.98 1048.20
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 64 a b c 200 400 600 800 1000 SE +/- 0.93, N = 3 SE +/- 0.28, N = 3 SE +/- 0.12, N = 3 952.20 951.88 951.85
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 64 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 64 a c 14 28 42 56 70 SE +/- 0.12, N = 3 SE +/- 0.07, N = 3 64.63 64.61
Intel TensorFlow Model: mobilenetv1_int8_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_int8_pretrained_model - Batch Size: 256 a c 700 1400 2100 2800 3500 SE +/- 2.91, N = 3 SE +/- 3.88, N = 3 3482.01 3482.91
Intel TensorFlow Model: mobilenetv1_fp32_pretrained_model - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: mobilenetv1_fp32_pretrained_model - Batch Size: 256 a b c 200 400 600 800 1000 SE +/- 0.47, N = 3 SE +/- 0.10, N = 3 SE +/- 1.00, N = 3 867.02 867.13 866.92
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 96 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 96 a c 150 300 450 600 750 SE +/- 0.52, N = 3 SE +/- 0.56, N = 3 691.01 690.86
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 16 a c 200 400 600 800 1000 SE +/- 1.60, N = 3 SE +/- 1.06, N = 3 953.78 953.58
AMD ZenDNN TensorFlow Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 256 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_mobilenetv1_1.0_imagenet_224_224_1.14G_1.1_Z4.0 - Batch Size: 256 a c 140 280 420 560 700 SE +/- 0.43, N = 3 SE +/- 0.46, N = 3 650.17 650.04
AMD ZenDNN TensorFlow Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better AMD ZenDNN TensorFlow 2.10 ZenDNN 4.0 Model: tf_inceptionv4_imagenet_299_299_24.55G_1.1_Z4.0 - Batch Size: 16 a c 14 28 42 56 70 SE +/- 0.01, N = 3 SE +/- 0.03, N = 3 62.78 62.79
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 16 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 16 a b c 70 140 210 280 350 SE +/- 5.23, N = 15 SE +/- 4.04, N = 15 SE +/- 5.26, N = 12 311.78 319.25 312.28
Intel TensorFlow Model: inceptionv4_int8_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_int8_pretrained_model - Batch Size: 1 a b c 20 40 60 80 100 SE +/- 0.54, N = 3 SE +/- 0.36, N = 3 SE +/- 3.79, N = 12 104.96 105.50 99.00
Intel TensorFlow Model: inceptionv4_fp32_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: inceptionv4_fp32_pretrained_model - Batch Size: 1 a b c 14 28 42 56 70 SE +/- 0.31, N = 3 SE +/- 0.59, N = 3 SE +/- 1.97, N = 15 61.65 61.00 57.42
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 1 OpenBenchmarking.org ms, Fewer Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 1 a b c 0.4469 0.8938 1.3407 1.7876 2.2345 SE +/- 0.019, N = 3 SE +/- 0.021, N = 3 SE +/- 0.041, N = 15 1.826 1.986 1.924
Intel TensorFlow Model: resnet50_int8_pretrained_model - Batch Size: 1 OpenBenchmarking.org images/sec, More Is Better Intel TensorFlow 2.12 Model: resnet50_int8_pretrained_model - Batch Size: 1 a b c 120 240 360 480 600 SE +/- 5.71, N = 3 SE +/- 5.31, N = 3 SE +/- 11.17, N = 15 547.78 503.50 523.17
Faiss Test: bench_polysemous_sift1m - Polysemous 30 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 30 a b c 0.1087 0.2174 0.3261 0.4348 0.5435 SE +/- 0.001, N = 3 SE +/- 0.003, N = 3 SE +/- 0.003, N = 3 0.482 0.481 0.483 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 34 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 34 a b c 0.1094 0.2188 0.3282 0.4376 0.547 SE +/- 0.000, N = 3 SE +/- 0.003, N = 3 SE +/- 0.003, N = 3 0.485 0.485 0.486 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 38 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 38 a b c 0.1116 0.2232 0.3348 0.4464 0.558 SE +/- 0.000, N = 3 SE +/- 0.003, N = 3 SE +/- 0.003, N = 3 0.495 0.494 0.496 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 42 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 42 a b c 0.1181 0.2362 0.3543 0.4724 0.5905 SE +/- 0.001, N = 3 SE +/- 0.003, N = 3 SE +/- 0.004, N = 3 0.525 0.522 0.525 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 46 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 46 a b c 0.1377 0.2754 0.4131 0.5508 0.6885 SE +/- 0.000, N = 3 SE +/- 0.003, N = 3 SE +/- 0.004, N = 3 0.612 0.610 0.612 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 50 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 50 a b c 0.187 0.374 0.561 0.748 0.935 SE +/- 0.001, N = 3 SE +/- 0.002, N = 3 SE +/- 0.005, N = 3 0.830 0.825 0.831 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 54 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 54 a b c 0.2891 0.5782 0.8673 1.1564 1.4455 SE +/- 0.000, N = 3 SE +/- 0.005, N = 3 SE +/- 0.008, N = 3 1.284 1.278 1.285 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 58 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 58 a b c 0.4802 0.9604 1.4406 1.9208 2.401 SE +/- 0.001, N = 3 SE +/- 0.006, N = 3 SE +/- 0.013, N = 3 2.132 2.126 2.134 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 62 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 62 a b c 0.7803 1.5606 2.3409 3.1212 3.9015 SE +/- 0.002, N = 3 SE +/- 0.002, N = 3 SE +/- 0.021, N = 3 3.466 3.449 3.468 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Faiss Test: bench_polysemous_sift1m - Polysemous 64 OpenBenchmarking.org ms per query, Fewer Is Better Faiss 1.7.4 Test: bench_polysemous_sift1m - Polysemous 64 a b c 0.9286 1.8572 2.7858 3.7144 4.643 SE +/- 0.002, N = 3 SE +/- 0.008, N = 3 SE +/- 0.024, N = 3 4.127 4.091 4.117 1. (F9X) gfortran options: -O2 -frecursive -m64 -fopenmp -msse3 -mssse3 -msse4.1 -mavx -mavx2 -fno-tree-vectorize -lm -lpthread -lgfortran -lc
Phoronix Test Suite v10.8.4