9950X onnx svt AMD Ryzen 9 9950X 16-Core testing with a ASUS ROG STRIX X670E-E GAMING WIFI (2204 BIOS) and AMD Radeon RX 7900 GRE 16GB on Ubuntu 24.04 via the Phoronix Test Suite.
HTML result view exported from: https://openbenchmarking.org/result/2408225-NE-9950XONNX44&grr&sor&rro .
9950X onnx svt Processor Motherboard Chipset Memory Disk Graphics Audio Monitor Network OS Kernel Desktop Display Server OpenGL Compiler File-System Screen Resolution a b c d e AMD Ryzen 9 9950X 16-Core @ 5.75GHz (16 Cores / 32 Threads) ASUS ROG STRIX X670E-E GAMING WIFI (2204 BIOS) AMD Device 14d8 2 x 32GB DDR5-6400MT/s Corsair CMK64GX5M2B6400C32 2000GB Corsair MP700 PRO AMD Radeon RX 7900 GRE 16GB AMD Navi 31 HDMI/DP DELL U2723QE Intel I225-V + Intel Wi-Fi 6E Ubuntu 24.04 6.10.0-phx (x86_64) GNOME Shell 46.0 X Server + Wayland 4.6 Mesa 24.2~git2406040600.8112d4~oibaf~n (git-8112d44 2024-06-04 noble-oibaf-ppa) (LLVM 17.0.6 DRM 3.57) GCC 13.2.0 ext4 3840x2160 OpenBenchmarking.org Kernel Details - Transparent Huge Pages: madvise Compiler Details - --build=x86_64-linux-gnu --disable-vtable-verify --disable-werror --enable-cet --enable-checking=release --enable-clocale=gnu --enable-default-pie --enable-gnu-unique-object --enable-languages=c,ada,c++,go,d,fortran,objc,obj-c++,m2 --enable-libphobos-checking=release --enable-libstdcxx-backtrace --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-multiarch --enable-multilib --enable-nls --enable-objc-gc=auto --enable-offload-defaulted --enable-offload-targets=nvptx-none=/build/gcc-13-uJ7kn6/gcc-13-13.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-uJ7kn6/gcc-13-13.2.0/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-default-libstdcxx-abi=new --with-gcc-major-version-only --with-multilib-list=m32,m64,mx32 --with-target-system-zlib=auto --with-tune=generic --without-cuda-driver -v Processor Details - Scaling Governor: amd-pstate-epp powersave (EPP: balance_performance) - CPU Microcode: 0xb40401a Python Details - Python 3.12.3 Security Details - gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: Not affected + reg_file_data_sampling: Not affected + retbleed: Not affected + spec_rstack_overflow: 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; STIBP: always-on; RSB filling; PBRSB-eIBRS: Not affected; BHI: Not affected + srbds: Not affected + tsx_async_abort: Not affected
9950X onnx svt svt-av1: Preset 3 - Beauty 4K 10-bit whisperfile: Medium svt-av1: Preset 3 - Bosphorus 4K svt-av1: Preset 5 - Beauty 4K 10-bit svt-av1: Preset 8 - Beauty 4K 10-bit onnx: yolov4 - CPU - Parallel onnx: yolov4 - CPU - Parallel whisperfile: Small svt-av1: Preset 3 - Bosphorus 1080p onnx: T5 Encoder - CPU - Standard onnx: T5 Encoder - CPU - Standard svt-av1: Preset 13 - Beauty 4K 10-bit onnx: GPT-2 - CPU - Parallel onnx: GPT-2 - CPU - Parallel onnx: GPT-2 - CPU - Standard onnx: GPT-2 - CPU - Standard onnx: bertsquad-12 - CPU - Parallel onnx: bertsquad-12 - CPU - Parallel onnx: ZFNet-512 - CPU - Parallel onnx: ZFNet-512 - CPU - Parallel onnx: ZFNet-512 - CPU - Standard onnx: ZFNet-512 - CPU - Standard onnx: yolov4 - CPU - Standard onnx: yolov4 - CPU - Standard onnx: T5 Encoder - CPU - Parallel onnx: T5 Encoder - CPU - Parallel svt-av1: Preset 5 - Bosphorus 4K onnx: ResNet101_DUC_HDC-12 - CPU - Parallel onnx: ResNet101_DUC_HDC-12 - CPU - Parallel onnx: ResNet101_DUC_HDC-12 - CPU - Standard onnx: ResNet101_DUC_HDC-12 - CPU - Standard onnx: fcn-resnet101-11 - CPU - Parallel onnx: fcn-resnet101-11 - CPU - Parallel onnx: fcn-resnet101-11 - CPU - Standard onnx: fcn-resnet101-11 - CPU - Standard onnx: bertsquad-12 - CPU - Standard onnx: bertsquad-12 - CPU - Standard onnx: ArcFace ResNet-100 - CPU - Parallel onnx: ArcFace ResNet-100 - CPU - Parallel onnx: ArcFace ResNet-100 - CPU - Standard onnx: ArcFace ResNet-100 - CPU - Standard onnx: Faster R-CNN R-50-FPN-int8 - CPU - Standard onnx: Faster R-CNN R-50-FPN-int8 - CPU - Standard onnx: Faster R-CNN R-50-FPN-int8 - CPU - Parallel onnx: Faster R-CNN R-50-FPN-int8 - CPU - Parallel onnx: CaffeNet 12-int8 - CPU - Parallel onnx: CaffeNet 12-int8 - CPU - Parallel onnx: CaffeNet 12-int8 - CPU - Standard onnx: CaffeNet 12-int8 - CPU - Standard onnx: ResNet50 v1-12-int8 - CPU - Parallel onnx: ResNet50 v1-12-int8 - CPU - Parallel onnx: ResNet50 v1-12-int8 - CPU - Standard onnx: ResNet50 v1-12-int8 - CPU - Standard onnx: super-resolution-10 - CPU - Parallel onnx: super-resolution-10 - CPU - Parallel onnx: super-resolution-10 - CPU - Standard onnx: super-resolution-10 - CPU - Standard svt-av1: Preset 5 - Bosphorus 1080p svt-av1: Preset 8 - Bosphorus 4K whisperfile: Tiny svt-av1: Preset 8 - Bosphorus 1080p svt-av1: Preset 13 - Bosphorus 4K svt-av1: Preset 13 - Bosphorus 1080p a b c d e 1.760 328.97184 11.814 7.574 10.593 129.644 7.71368 127.97316 37.313 5.24802 190.624 19.400 6.62484 150.865 6.19550 161.414 84.9286 11.7750 14.0673 71.0823 7.86110 127.185 72.5475 13.7840 5.24748 190.549 44.104 705.209 1.41808 390.73 2.55959 449.481 2.22498 201.866 4.95375 39.8348 25.1022 40.8495 24.4794 17.9386 55.7411 16.2609 61.4985 24.1918 41.3330 4.07620 245.278 1.17405 851.244 6.55079 152.641 1.72670 578.940 4.50538 221.936 4.44116 225.156 128.707 98.657 29.63488 302.419 257.201 1009.640 1.758 327.12841 11.821 7.588 10.609 130.124 7.68899 128.23627 37.209 5.18948 192.676 19.404 6.60878 151.245 6.37911 156.717 86.1980 11.6016 14.1239 70.7958 7.95942 125.625 72.0326 13.8825 5.24914 190.482 43.974 708.821 1.41079 391.384 2.55502 444.436 2.25003 202.881 4.92894 39.9139 25.0524 40.9041 24.4461 17.8392 56.0511 16.5315 60.4839 24.2099 41.3018 4.1004 243.727 1.17235 852.418 6.54126 152.853 1.7433 573.425 4.53509 220.47 4.4613 224.14 127.961 99.255 30.31093 304.327 258.647 1011.236 1.76 329.23828 11.794 7.583 10.527 124.41 8.03773 129.62575 37.198 5.18564 192.823 19.419 6.6827 149.566 6.34248 157.628 83.4468 11.9832 13.9657 71.5953 7.93481 125.997 71.8677 13.914 5.32595 187.735 43.769 703.891 1.42067 387.873 2.57815 444.847 2.24795 200.78 4.98053 39.924 25.0458 41.1806 24.2813 17.8517 56.0119 16.076 62.198 24.4512 40.8941 4.12702 242.238 1.16879 855.082 6.68702 149.511 1.74974 571.193 4.66209 214.475 4.46838 223.785 128.368 99.222 30.03969 303.057 257.418 1012.768 1.759 328.28944 11.818 7.54 10.525 115.894 8.62837 129.79648 37.151 5.46322 183.018 19.488 6.65046 150.289 6.33978 157.689 85.4013 11.7088 14.1704 70.5538 7.86194 127.169 71.6925 13.948 5.2459 190.604 43.874 709.79 1.40886 390.94 2.55793 447.063 2.2368 202.724 4.93277 42.7142 23.4098 40.3648 24.7729 17.8447 56.034 16.0884 62.1495 24.3388 41.0832 4.06632 245.867 1.17003 854.139 6.69723 149.296 1.74031 574.317 4.4796 223.209 4.45929 224.24 128.103 98.242 30.03582 303.122 255.705 1011.433 1.759 328.01325 11.744 7.541 10.737 127.801 7.82444 129.38091 37.197 5.17395 193.253 19.487 6.61099 151.197 6.34139 157.65 85.0842 11.7526 13.9577 71.6359 7.93399 126.016 72.9637 13.705 5.22498 191.361 43.739 695.937 1.43691 392.057 2.55063 446.891 2.23767 202.807 4.93075 39.9589 25.0243 41.2192 24.2594 17.8852 55.9068 16.578 60.315 24.4935 40.8234 4.1821 239.032 1.15972 861.654 6.66991 149.907 1.7372 575.415 4.58833 217.917 4.54187 220.163 127.605 98.06 30.37277 301.639 256.152 1010.763 OpenBenchmarking.org
SVT-AV1 Encoder Mode: Preset 3 - Input: Beauty 4K 10-bit OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 3 - Input: Beauty 4K 10-bit b d e a c 0.396 0.792 1.188 1.584 1.98 SE +/- 0.003, N = 3 SE +/- 0.005, N = 3 1.758 1.759 1.759 1.760 1.760 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Whisperfile Model Size: Medium OpenBenchmarking.org Seconds, Fewer Is Better Whisperfile 20Aug24 Model Size: Medium c a d e b 70 140 210 280 350 SE +/- 0.20, N = 3 329.24 328.97 328.29 328.01 327.13
SVT-AV1 Encoder Mode: Preset 3 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 3 - Input: Bosphorus 4K e c a d b 3 6 9 12 15 SE +/- 0.04, N = 3 SE +/- 0.00, N = 3 11.74 11.79 11.81 11.82 11.82 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
SVT-AV1 Encoder Mode: Preset 5 - Input: Beauty 4K 10-bit OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 5 - Input: Beauty 4K 10-bit d e a c b 2 4 6 8 10 SE +/- 0.022, N = 3 SE +/- 0.052, N = 3 7.540 7.541 7.574 7.583 7.588 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
SVT-AV1 Encoder Mode: Preset 8 - Input: Beauty 4K 10-bit OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 8 - Input: Beauty 4K 10-bit d c a b e 3 6 9 12 15 SE +/- 0.03, N = 3 SE +/- 0.04, N = 3 10.53 10.53 10.59 10.61 10.74 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
ONNX Runtime Model: yolov4 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: yolov4 - Device: CPU - Executor: Parallel b a e c d 30 60 90 120 150 SE +/- 1.08, N = 9 SE +/- 0.73, N = 3 130.12 129.64 127.80 124.41 115.89 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: yolov4 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: yolov4 - Device: CPU - Executor: Parallel b a e c d 2 4 6 8 10 SE +/- 0.06383, N = 9 SE +/- 0.04367, N = 3 7.68899 7.71368 7.82444 8.03773 8.62837 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
Whisperfile Model Size: Small OpenBenchmarking.org Seconds, Fewer Is Better Whisperfile 20Aug24 Model Size: Small d c e b a 30 60 90 120 150 SE +/- 0.21, N = 3 129.80 129.63 129.38 128.24 127.97
SVT-AV1 Encoder Mode: Preset 3 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 3 - Input: Bosphorus 1080p d e c b a 9 18 27 36 45 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 37.15 37.20 37.20 37.21 37.31 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
ONNX Runtime Model: T5 Encoder - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: T5 Encoder - Device: CPU - Executor: Standard d a b c e 1.2292 2.4584 3.6876 4.9168 6.146 SE +/- 0.05529, N = 6 SE +/- 0.00692, N = 3 5.46322 5.24802 5.18948 5.18564 5.17395 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: T5 Encoder - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: T5 Encoder - Device: CPU - Executor: Standard d a b c e 40 80 120 160 200 SE +/- 1.93, N = 6 SE +/- 0.25, N = 3 183.02 190.62 192.68 192.82 193.25 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
SVT-AV1 Encoder Mode: Preset 13 - Input: Beauty 4K 10-bit OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 13 - Input: Beauty 4K 10-bit a b c e d 5 10 15 20 25 SE +/- 0.01, N = 3 SE +/- 0.01, N = 3 19.40 19.40 19.42 19.49 19.49 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
ONNX Runtime Model: GPT-2 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: GPT-2 - Device: CPU - Executor: Parallel c d a e b 2 4 6 8 10 SE +/- 0.00827, N = 3 SE +/- 0.02289, N = 3 6.68270 6.65046 6.62484 6.61099 6.60878 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: GPT-2 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: GPT-2 - Device: CPU - Executor: Parallel c d a e b 30 60 90 120 150 SE +/- 0.19, N = 3 SE +/- 0.53, N = 3 149.57 150.29 150.87 151.20 151.25 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: GPT-2 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: GPT-2 - Device: CPU - Executor: Standard b c e d a 2 4 6 8 10 SE +/- 0.01653, N = 3 SE +/- 0.07694, N = 3 6.37911 6.34248 6.34139 6.33978 6.19550 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: GPT-2 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: GPT-2 - Device: CPU - Executor: Standard b c e d a 40 80 120 160 200 SE +/- 0.41, N = 3 SE +/- 1.98, N = 3 156.72 157.63 157.65 157.69 161.41 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: bertsquad-12 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: bertsquad-12 - Device: CPU - Executor: Parallel b d e a c 20 40 60 80 100 SE +/- 0.50, N = 3 SE +/- 0.50, N = 3 86.20 85.40 85.08 84.93 83.45 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: bertsquad-12 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: bertsquad-12 - Device: CPU - Executor: Parallel b d e a c 3 6 9 12 15 SE +/- 0.07, N = 3 SE +/- 0.07, N = 3 11.60 11.71 11.75 11.78 11.98 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ZFNet-512 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ZFNet-512 - Device: CPU - Executor: Parallel d b a c e 4 8 12 16 20 SE +/- 0.07, N = 3 SE +/- 0.08, N = 3 14.17 14.12 14.07 13.97 13.96 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ZFNet-512 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ZFNet-512 - Device: CPU - Executor: Parallel d b a c e 16 32 48 64 80 SE +/- 0.36, N = 3 SE +/- 0.42, N = 3 70.55 70.80 71.08 71.60 71.64 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ZFNet-512 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ZFNet-512 - Device: CPU - Executor: Standard b c e d a 2 4 6 8 10 SE +/- 0.06114, N = 3 SE +/- 0.03518, N = 3 7.95942 7.93481 7.93399 7.86194 7.86110 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ZFNet-512 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ZFNet-512 - Device: CPU - Executor: Standard b c e d a 30 60 90 120 150 SE +/- 0.96, N = 3 SE +/- 0.57, N = 3 125.63 126.00 126.02 127.17 127.19 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: yolov4 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: yolov4 - Device: CPU - Executor: Standard e a b c d 16 32 48 64 80 SE +/- 0.26, N = 3 SE +/- 0.27, N = 3 72.96 72.55 72.03 71.87 71.69 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: yolov4 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: yolov4 - Device: CPU - Executor: Standard e a b c d 4 8 12 16 20 SE +/- 0.05, N = 3 SE +/- 0.05, N = 3 13.71 13.78 13.88 13.91 13.95 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: T5 Encoder - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: T5 Encoder - Device: CPU - Executor: Parallel c b a d e 1.1983 2.3966 3.5949 4.7932 5.9915 SE +/- 0.01022, N = 3 SE +/- 0.02621, N = 3 5.32595 5.24914 5.24748 5.24590 5.22498 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: T5 Encoder - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: T5 Encoder - Device: CPU - Executor: Parallel c b a d e 40 80 120 160 200 SE +/- 0.37, N = 3 SE +/- 0.96, N = 3 187.74 190.48 190.55 190.60 191.36 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
SVT-AV1 Encoder Mode: Preset 5 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 5 - Input: Bosphorus 4K e c d b a 10 20 30 40 50 SE +/- 0.13, N = 3 SE +/- 0.06, N = 3 43.74 43.77 43.87 43.97 44.10 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
ONNX Runtime Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Parallel d b a c e 150 300 450 600 750 SE +/- 3.47, N = 3 709.79 708.82 705.21 703.89 695.94 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Parallel d b a c e 0.3233 0.6466 0.9699 1.2932 1.6165 SE +/- 0.00694, N = 3 1.40886 1.41079 1.41808 1.42067 1.43691 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Standard e b d a c 90 180 270 360 450 SE +/- 3.03, N = 3 392.06 391.38 390.94 390.73 387.87 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ResNet101_DUC_HDC-12 - Device: CPU - Executor: Standard e b d a c 0.5801 1.1602 1.7403 2.3204 2.9005 SE +/- 0.02001, N = 3 2.55063 2.55502 2.55793 2.55959 2.57815 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: fcn-resnet101-11 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: fcn-resnet101-11 - Device: CPU - Executor: Parallel a d e c b 100 200 300 400 500 SE +/- 3.07, N = 3 449.48 447.06 446.89 444.85 444.44 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: fcn-resnet101-11 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: fcn-resnet101-11 - Device: CPU - Executor: Parallel a d e c b 0.5063 1.0126 1.5189 2.0252 2.5315 SE +/- 0.01511, N = 3 2.22498 2.23680 2.23767 2.24795 2.25003 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: fcn-resnet101-11 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: fcn-resnet101-11 - Device: CPU - Executor: Standard b e d a c 40 80 120 160 200 SE +/- 0.28, N = 3 202.88 202.81 202.72 201.87 200.78 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: fcn-resnet101-11 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: fcn-resnet101-11 - Device: CPU - Executor: Standard b e d a c 1.1206 2.2412 3.3618 4.4824 5.603 SE +/- 0.00675, N = 3 4.92894 4.93075 4.93277 4.95375 4.98053 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: bertsquad-12 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: bertsquad-12 - Device: CPU - Executor: Standard d e c b a 10 20 30 40 50 SE +/- 0.07, N = 3 42.71 39.96 39.92 39.91 39.83 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: bertsquad-12 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: bertsquad-12 - Device: CPU - Executor: Standard d e c b a 6 12 18 24 30 SE +/- 0.04, N = 3 23.41 25.02 25.05 25.05 25.10 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ArcFace ResNet-100 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ArcFace ResNet-100 - Device: CPU - Executor: Parallel e c b a d 9 18 27 36 45 SE +/- 0.15, N = 3 41.22 41.18 40.90 40.85 40.36 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ArcFace ResNet-100 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ArcFace ResNet-100 - Device: CPU - Executor: Parallel e c b a d 6 12 18 24 30 SE +/- 0.09, N = 3 24.26 24.28 24.45 24.48 24.77 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard a e c d b 4 8 12 16 20 SE +/- 0.04, N = 3 17.94 17.89 17.85 17.84 17.84 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ArcFace ResNet-100 - Device: CPU - Executor: Standard a e c d b 13 26 39 52 65 SE +/- 0.11, N = 3 55.74 55.91 56.01 56.03 56.05 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Standard e b a d c 4 8 12 16 20 SE +/- 0.13, N = 3 16.58 16.53 16.26 16.09 16.08 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Standard e b a d c 14 28 42 56 70 SE +/- 0.49, N = 3 60.32 60.48 61.50 62.15 62.20 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Parallel e c d b a 6 12 18 24 30 SE +/- 0.06, N = 3 24.49 24.45 24.34 24.21 24.19 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: Faster R-CNN R-50-FPN-int8 - Device: CPU - Executor: Parallel e c d b a 9 18 27 36 45 SE +/- 0.10, N = 3 40.82 40.89 41.08 41.30 41.33 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: CaffeNet 12-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: CaffeNet 12-int8 - Device: CPU - Executor: Parallel e c b a d 0.941 1.882 2.823 3.764 4.705 SE +/- 0.03127, N = 3 4.18210 4.12702 4.10040 4.07620 4.06632 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: CaffeNet 12-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: CaffeNet 12-int8 - Device: CPU - Executor: Parallel e c b a d 50 100 150 200 250 SE +/- 1.89, N = 3 239.03 242.24 243.73 245.28 245.87 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: CaffeNet 12-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: CaffeNet 12-int8 - Device: CPU - Executor: Standard a b d c e 0.2642 0.5284 0.7926 1.0568 1.321 SE +/- 0.00364, N = 3 1.17405 1.17235 1.17003 1.16879 1.15972 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: CaffeNet 12-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: CaffeNet 12-int8 - Device: CPU - Executor: Standard a b d c e 200 400 600 800 1000 SE +/- 2.64, N = 3 851.24 852.42 854.14 855.08 861.65 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Parallel d c e a b 2 4 6 8 10 SE +/- 0.03664, N = 3 6.69723 6.68702 6.66991 6.55079 6.54126 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Parallel d c e a b 30 60 90 120 150 SE +/- 0.85, N = 3 149.30 149.51 149.91 152.64 152.85 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Standard c b d e a 0.3937 0.7874 1.1811 1.5748 1.9685 SE +/- 0.00641, N = 3 1.74974 1.74330 1.74031 1.73720 1.72670 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: ResNet50 v1-12-int8 - Device: CPU - Executor: Standard c b d e a 130 260 390 520 650 SE +/- 2.14, N = 3 571.19 573.43 574.32 575.42 578.94 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: super-resolution-10 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: super-resolution-10 - Device: CPU - Executor: Parallel c e b a d 1.049 2.098 3.147 4.196 5.245 SE +/- 0.02467, N = 3 4.66209 4.58833 4.53509 4.50538 4.47960 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: super-resolution-10 - Device: CPU - Executor: Parallel OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: super-resolution-10 - Device: CPU - Executor: Parallel c e b a d 50 100 150 200 250 SE +/- 1.21, N = 3 214.48 217.92 220.47 221.94 223.21 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: super-resolution-10 - Device: CPU - Executor: Standard OpenBenchmarking.org Inference Time Cost (ms), Fewer Is Better ONNX Runtime 1.19 Model: super-resolution-10 - Device: CPU - Executor: Standard e c b d a 1.0219 2.0438 3.0657 4.0876 5.1095 SE +/- 0.00178, N = 3 4.54187 4.46838 4.46130 4.45929 4.44116 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
ONNX Runtime Model: super-resolution-10 - Device: CPU - Executor: Standard OpenBenchmarking.org Inferences Per Second, More Is Better ONNX Runtime 1.19 Model: super-resolution-10 - Device: CPU - Executor: Standard e c b d a 50 100 150 200 250 SE +/- 0.09, N = 3 220.16 223.79 224.14 224.24 225.16 1. (CXX) g++ options: -O3 -march=native -ffunction-sections -fdata-sections -mtune=native -flto=auto -fno-fat-lto-objects -ldl -lrt
SVT-AV1 Encoder Mode: Preset 5 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 5 - Input: Bosphorus 1080p e b d c a 30 60 90 120 150 SE +/- 0.17, N = 3 SE +/- 0.11, N = 3 127.61 127.96 128.10 128.37 128.71 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
SVT-AV1 Encoder Mode: Preset 8 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 8 - Input: Bosphorus 4K e d a c b 20 40 60 80 100 SE +/- 0.16, N = 3 SE +/- 0.23, N = 3 98.06 98.24 98.66 99.22 99.26 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Whisperfile Model Size: Tiny OpenBenchmarking.org Seconds, Fewer Is Better Whisperfile 20Aug24 Model Size: Tiny e b c d a 7 14 21 28 35 SE +/- 0.12, N = 3 30.37 30.31 30.04 30.04 29.63
SVT-AV1 Encoder Mode: Preset 8 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 8 - Input: Bosphorus 1080p e a c d b 70 140 210 280 350 SE +/- 0.64, N = 3 SE +/- 0.56, N = 3 301.64 302.42 303.06 303.12 304.33 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
SVT-AV1 Encoder Mode: Preset 13 - Input: Bosphorus 4K OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 13 - Input: Bosphorus 4K d e a c b 60 120 180 240 300 SE +/- 0.27, N = 3 SE +/- 0.18, N = 3 255.71 256.15 257.20 257.42 258.65 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
SVT-AV1 Encoder Mode: Preset 13 - Input: Bosphorus 1080p OpenBenchmarking.org Frames Per Second, More Is Better SVT-AV1 2.2 Encoder Mode: Preset 13 - Input: Bosphorus 1080p a e b d c 200 400 600 800 1000 SE +/- 2.29, N = 3 SE +/- 1.10, N = 3 1009.64 1010.76 1011.24 1011.43 1012.77 1. (CXX) g++ options: -march=native -mno-avx -mavx2 -mavx512f -mavx512bw -mavx512dq
Phoronix Test Suite v10.8.5