a Benchmarks for a future article.
Compare your own system(s) to this result file with the
Phoronix Test Suite by running the command:
phoronix-test-suite benchmark 2312143-NE-A8154652071 a Kernel Notes: Transparent Huge Pages: madviseProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
b c d Processor: 2 x INTEL XEON PLATINUM 8592+ @ 3.90GHz (128 Cores / 256 Threads), Motherboard: Quanta Cloud S6Q-MB-MPS (3B05.TEL4P1 BIOS), Chipset: Intel Device 1bce, Memory: 1008GB, Disk: 3201GB Micron_7450_MTFDKCB3T2TFS, Graphics: ASPEED, Network: 2 x Intel X710 for 10GBASE-T
OS: Ubuntu 23.10, Kernel: 6.5.0-13-generic (x86_64), Compiler: GCC 13.2.0, File-System: ext4, Screen Resolution: 1920x1080
Kernel Notes: Transparent Huge Pages: madviseCompiler 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,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-defaulted --enable-offload-targets=nvptx-none=/build/gcc-13-XYspKM/gcc-13-13.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-XYspKM/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-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 -vProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
a OpenBenchmarking.org Phoronix Test Suite 2 x INTEL XEON PLATINUM 8592+ @ 3.90GHz (128 Cores / 256 Threads) Quanta Cloud S6Q-MB-MPS (3B05.TEL4P1 BIOS) Intel Device 1bce 1008GB 3201GB Micron_7450_MTFDKCB3T2TFS ASPEED 2 x Intel X710 for 10GBASE-T Ubuntu 23.10 6.5.0-13-generic (x86_64) GCC 13.2.0 ext4 1920x1080 Processor Motherboard Chipset Memory Disk Graphics Network OS Kernel Compiler File-System Screen Resolution A Benchmarks System Logs - Transparent Huge Pages: madvise - Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161 - Python 3.11.6 - gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected - d: --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,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-defaulted --enable-offload-targets=nvptx-none=/build/gcc-13-XYspKM/gcc-13-13.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-XYspKM/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-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
a b c d Result Overview Phoronix Test Suite 100% 100% 101% 101% 102% NWChem Neural Magic DeepSparse WRF
a deepsparse: NLP Document Classification, oBERT base uncased on IMDB - Asynchronous Multi-Stream deepsparse: NLP Document Classification, oBERT base uncased on IMDB - Asynchronous Multi-Stream deepsparse: NLP Document Classification, oBERT base uncased on IMDB - Synchronous Single-Stream deepsparse: NLP Document Classification, oBERT base uncased on IMDB - Synchronous Single-Stream deepsparse: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Asynchronous Multi-Stream deepsparse: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Asynchronous Multi-Stream deepsparse: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Synchronous Single-Stream deepsparse: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Synchronous Single-Stream deepsparse: ResNet-50, Baseline - Asynchronous Multi-Stream deepsparse: ResNet-50, Baseline - Asynchronous Multi-Stream deepsparse: ResNet-50, Baseline - Synchronous Single-Stream deepsparse: ResNet-50, Baseline - Synchronous Single-Stream deepsparse: ResNet-50, Sparse INT8 - Asynchronous Multi-Stream deepsparse: ResNet-50, Sparse INT8 - Asynchronous Multi-Stream deepsparse: ResNet-50, Sparse INT8 - Synchronous Single-Stream deepsparse: ResNet-50, Sparse INT8 - Synchronous Single-Stream deepsparse: CV Detection, YOLOv5s COCO - Asynchronous Multi-Stream deepsparse: CV Detection, YOLOv5s COCO - Asynchronous Multi-Stream deepsparse: CV Detection, YOLOv5s COCO - Synchronous Single-Stream deepsparse: CV Detection, YOLOv5s COCO - Synchronous Single-Stream deepsparse: BERT-Large, NLP Question Answering - Asynchronous Multi-Stream deepsparse: BERT-Large, NLP Question Answering - Asynchronous Multi-Stream deepsparse: BERT-Large, NLP Question Answering - Synchronous Single-Stream deepsparse: BERT-Large, NLP Question Answering - Synchronous Single-Stream deepsparse: CV Classification, ResNet-50 ImageNet - Asynchronous Multi-Stream deepsparse: CV Classification, ResNet-50 ImageNet - Asynchronous Multi-Stream deepsparse: CV Classification, ResNet-50 ImageNet - Synchronous Single-Stream deepsparse: CV Classification, ResNet-50 ImageNet - Synchronous Single-Stream deepsparse: CV Detection, YOLOv5s COCO, Sparse INT8 - Asynchronous Multi-Stream deepsparse: CV Detection, YOLOv5s COCO, Sparse INT8 - Asynchronous Multi-Stream deepsparse: CV Detection, YOLOv5s COCO, Sparse INT8 - Synchronous Single-Stream deepsparse: CV Detection, YOLOv5s COCO, Sparse INT8 - Synchronous Single-Stream deepsparse: NLP Text Classification, DistilBERT mnli - Asynchronous Multi-Stream deepsparse: NLP Text Classification, DistilBERT mnli - Asynchronous Multi-Stream deepsparse: NLP Text Classification, DistilBERT mnli - Synchronous Single-Stream deepsparse: NLP Text Classification, DistilBERT mnli - Synchronous Single-Stream deepsparse: CV Segmentation, 90% Pruned YOLACT Pruned - Asynchronous Multi-Stream deepsparse: CV Segmentation, 90% Pruned YOLACT Pruned - Asynchronous Multi-Stream deepsparse: CV Segmentation, 90% Pruned YOLACT Pruned - Synchronous Single-Stream deepsparse: CV Segmentation, 90% Pruned YOLACT Pruned - Synchronous Single-Stream deepsparse: BERT-Large, NLP Question Answering, Sparse INT8 - Asynchronous Multi-Stream deepsparse: BERT-Large, NLP Question Answering, Sparse INT8 - Asynchronous Multi-Stream deepsparse: BERT-Large, NLP Question Answering, Sparse INT8 - Synchronous Single-Stream deepsparse: BERT-Large, NLP Question Answering, Sparse INT8 - Synchronous Single-Stream deepsparse: NLP Token Classification, BERT base uncased conll2003 - Asynchronous Multi-Stream deepsparse: NLP Token Classification, BERT base uncased conll2003 - Asynchronous Multi-Stream deepsparse: NLP Token Classification, BERT base uncased conll2003 - Synchronous Single-Stream deepsparse: NLP Token Classification, BERT base uncased conll2003 - Synchronous Single-Stream nwchem: C240 Buckyball wrf: conus 2.5km a b c d 133.5159 475.6964 31.1945 32.0613 4554.9796 14.0362 238.6523 4.2083 1829.9644 34.9169 338.9562 2.9492 11228.6625 5.6852 1030.3181 0.9681 828.0384 77.1758 190.8420 5.2382 156.0563 407.2777 29.9421 33.4022 1832.7927 34.8681 339.0416 2.9484 854.0525 74.8177 194.2476 5.1485 1231.2049 51.9301 199.6930 5.0073 183.4966 347.6637 34.9620 28.5876 1880.7688 33.9808 96.4300 10.3634 133.6023 475.1265 31.6954 31.5821 1744 5566.729 133.7189 474.1852 31.2336 32.0368 4563.5097 14.0081 237.0460 4.2295 1824.3759 35.0063 337.6799 2.9602 11169.8517 5.7140 1027.5117 0.9706 828.6532 77.0924 190.0399 5.2617 155.7222 408.6263 30.3891 32.8989 1817.9939 35.1423 338.8583 2.9500 852.9455 74.9435 193.5972 5.1659 1231.2939 51.9273 199.3380 5.0196 180.3157 353.6889 35.1065 28.4701 1875.9035 34.0815 96.3790 10.3695 133.8239 474.3687 31.5332 31.7303 1730.7 5600.976 133.4843 475.6804 31.6300 31.6501 4563.4278 14.0090 230.7640 4.3436 1827.9780 34.9683 338.5536 2.9530 11163.6593 5.7165 1043.2804 0.9556 828.4462 77.1235 189.8175 5.2651 155.7677 408.9699 30.2944 32.9984 1834.4771 34.8217 338.2192 2.9555 854.2521 74.7945 193.0872 5.1802 1231.2736 51.9192 199.3323 5.0179 177.1512 359.8919 35.0057 28.5521 1879.8388 33.9939 96.5474 10.3507 133.6004 475.3497 31.7202 31.5618 1757.3 5583.112 133.0079 477.6108 32.8936 30.4222 4556.0233 14.0315 240.2336 4.1604 1837.4977 34.7567 338.1466 2.9573 11221.1357 5.6908 1032.2410 0.9667 825.0841 77.4573 190.1376 5.2589 155.6020 408.1638 29.5231 33.8635 1819.6861 35.1013 339.3671 2.9455 850.8926 75.0848 193.8186 5.1614 1226.7960 52.0706 196.7874 5.0808 181.2029 352.2235 34.9987 28.5586 1865.0193 34.2508 95.6981 10.4501 133.5969 475.4464 33.0540 30.2891 1748 5617.197 OpenBenchmarking.org
Neural Magic DeepSparse This is a benchmark of Neural Magic's DeepSparse using its built-in deepsparse.benchmark utility and various models from their SparseZoo (https://sparsezoo.neuralmagic.com/). Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream d c b a 30 60 90 120 150 SE +/- 0.11, N = 3 SE +/- 0.16, N = 3 SE +/- 0.13, N = 3 SE +/- 0.15, N = 3 133.01 133.48 133.72 133.52
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Asynchronous Multi-Stream d c b a 100 200 300 400 500 SE +/- 0.44, N = 3 SE +/- 0.52, N = 3 SE +/- 0.47, N = 3 SE +/- 0.61, N = 3 477.61 475.68 474.19 475.70
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Synchronous Single-Stream d c b a 8 16 24 32 40 SE +/- 0.26, N = 15 SE +/- 0.32, N = 15 SE +/- 0.25, N = 15 SE +/- 0.38, N = 4 32.89 31.63 31.23 31.19
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Document Classification, oBERT base uncased on IMDB - Scenario: Synchronous Single-Stream d c b a 7 14 21 28 35 SE +/- 0.25, N = 15 SE +/- 0.32, N = 15 SE +/- 0.27, N = 15 SE +/- 0.40, N = 4 30.42 31.65 32.04 32.06
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 1000 2000 3000 4000 5000 SE +/- 42.83, N = 6 SE +/- 43.60, N = 6 SE +/- 40.39, N = 7 SE +/- 49.40, N = 5 4556.02 4563.43 4563.51 4554.98
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 4 8 12 16 20 SE +/- 0.14, N = 6 SE +/- 0.14, N = 6 SE +/- 0.13, N = 7 SE +/- 0.16, N = 5 14.03 14.01 14.01 14.04
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 50 100 150 200 250 SE +/- 3.16, N = 3 SE +/- 3.66, N = 15 SE +/- 3.82, N = 15 SE +/- 4.61, N = 15 240.23 230.76 237.05 238.65
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, BERT base uncased SST2, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 0.9773 1.9546 2.9319 3.9092 4.8865 SE +/- 0.0540, N = 3 SE +/- 0.0661, N = 15 SE +/- 0.0666, N = 15 SE +/- 0.0821, N = 15 4.1604 4.3436 4.2295 4.2083
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Baseline - Scenario: Asynchronous Multi-Stream d c b a 400 800 1200 1600 2000 SE +/- 14.66, N = 3 SE +/- 17.94, N = 3 SE +/- 13.19, N = 3 SE +/- 4.85, N = 3 1837.50 1827.98 1824.38 1829.96
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Baseline - Scenario: Asynchronous Multi-Stream d c b a 8 16 24 32 40 SE +/- 0.29, N = 3 SE +/- 0.35, N = 3 SE +/- 0.25, N = 3 SE +/- 0.08, N = 3 34.76 34.97 35.01 34.92
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Baseline - Scenario: Synchronous Single-Stream d c b a 70 140 210 280 350 SE +/- 2.94, N = 12 SE +/- 2.56, N = 12 SE +/- 3.44, N = 6 SE +/- 2.76, N = 9 338.15 338.55 337.68 338.96
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Baseline - Scenario: Synchronous Single-Stream d c b a 0.666 1.332 1.998 2.664 3.33 SE +/- 0.0282, N = 12 SE +/- 0.0242, N = 12 SE +/- 0.0315, N = 6 SE +/- 0.0255, N = 9 2.9573 2.9530 2.9602 2.9492
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 2K 4K 6K 8K 10K SE +/- 77.36, N = 12 SE +/- 101.43, N = 7 SE +/- 89.80, N = 9 SE +/- 71.77, N = 13 11221.14 11163.66 11169.85 11228.66
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 1.2862 2.5724 3.8586 5.1448 6.431 SE +/- 0.0425, N = 12 SE +/- 0.0544, N = 7 SE +/- 0.0484, N = 9 SE +/- 0.0393, N = 13 5.6908 5.7165 5.7140 5.6852
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 200 400 600 800 1000 SE +/- 9.64, N = 7 SE +/- 5.37, N = 3 SE +/- 9.47, N = 3 SE +/- 11.12, N = 4 1032.24 1043.28 1027.51 1030.32
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: ResNet-50, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 0.2184 0.4368 0.6552 0.8736 1.092 SE +/- 0.0094, N = 7 SE +/- 0.0048, N = 3 SE +/- 0.0090, N = 3 SE +/- 0.0105, N = 4 0.9667 0.9556 0.9706 0.9681
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO - Scenario: Asynchronous Multi-Stream d c b a 200 400 600 800 1000 SE +/- 9.22, N = 3 SE +/- 6.81, N = 3 SE +/- 6.37, N = 3 SE +/- 7.51, N = 3 825.08 828.45 828.65 828.04
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO - Scenario: Asynchronous Multi-Stream d c b a 20 40 60 80 100 SE +/- 0.87, N = 3 SE +/- 0.64, N = 3 SE +/- 0.63, N = 3 SE +/- 0.70, N = 3 77.46 77.12 77.09 77.18
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO - Scenario: Synchronous Single-Stream d c b a 40 80 120 160 200 SE +/- 1.52, N = 12 SE +/- 1.69, N = 7 SE +/- 1.73, N = 12 SE +/- 1.29, N = 12 190.14 189.82 190.04 190.84
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO - Scenario: Synchronous Single-Stream d c b a 1.1846 2.3692 3.5538 4.7384 5.923 SE +/- 0.0457, N = 12 SE +/- 0.0492, N = 7 SE +/- 0.0526, N = 12 SE +/- 0.0382, N = 12 5.2589 5.2651 5.2617 5.2382
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering - Scenario: Asynchronous Multi-Stream d c b a 30 60 90 120 150 SE +/- 1.10, N = 3 SE +/- 1.19, N = 3 SE +/- 0.66, N = 3 SE +/- 0.66, N = 3 155.60 155.77 155.72 156.06
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering - Scenario: Asynchronous Multi-Stream d c b a 90 180 270 360 450 SE +/- 2.06, N = 3 SE +/- 2.55, N = 3 SE +/- 1.91, N = 3 SE +/- 1.62, N = 3 408.16 408.97 408.63 407.28
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering - Scenario: Synchronous Single-Stream d c b a 7 14 21 28 35 SE +/- 0.10, N = 3 SE +/- 0.15, N = 3 SE +/- 0.27, N = 3 SE +/- 0.30, N = 6 29.52 30.29 30.39 29.94
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering - Scenario: Synchronous Single-Stream d c b a 8 16 24 32 40 SE +/- 0.12, N = 3 SE +/- 0.16, N = 3 SE +/- 0.30, N = 3 SE +/- 0.34, N = 6 33.86 33.00 32.90 33.40
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream d c b a 400 800 1200 1600 2000 SE +/- 21.20, N = 3 SE +/- 5.61, N = 3 SE +/- 7.71, N = 3 SE +/- 22.80, N = 3 1819.69 1834.48 1817.99 1832.79
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Classification, ResNet-50 ImageNet - Scenario: Asynchronous Multi-Stream d c b a 8 16 24 32 40 SE +/- 0.39, N = 3 SE +/- 0.11, N = 3 SE +/- 0.14, N = 3 SE +/- 0.43, N = 3 35.10 34.82 35.14 34.87
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Classification, ResNet-50 ImageNet - Scenario: Synchronous Single-Stream d c b a 70 140 210 280 350 SE +/- 2.24, N = 12 SE +/- 3.30, N = 6 SE +/- 2.82, N = 9 SE +/- 2.66, N = 10 339.37 338.22 338.86 339.04
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Classification, ResNet-50 ImageNet - Scenario: Synchronous Single-Stream d c b a 0.665 1.33 1.995 2.66 3.325 SE +/- 0.0208, N = 12 SE +/- 0.0300, N = 6 SE +/- 0.0260, N = 9 SE +/- 0.0247, N = 10 2.9455 2.9555 2.9500 2.9484
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 200 400 600 800 1000 SE +/- 10.96, N = 3 SE +/- 9.10, N = 3 SE +/- 10.26, N = 3 SE +/- 8.70, N = 3 850.89 854.25 852.95 854.05
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 20 40 60 80 100 SE +/- 0.97, N = 3 SE +/- 0.80, N = 3 SE +/- 0.92, N = 3 SE +/- 0.77, N = 3 75.08 74.79 74.94 74.82
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 40 80 120 160 200 SE +/- 1.62, N = 12 SE +/- 1.70, N = 12 SE +/- 1.53, N = 13 SE +/- 1.36, N = 12 193.82 193.09 193.60 194.25
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Detection, YOLOv5s COCO, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 1.1655 2.331 3.4965 4.662 5.8275 SE +/- 0.0471, N = 12 SE +/- 0.0500, N = 12 SE +/- 0.0445, N = 13 SE +/- 0.0389, N = 12 5.1614 5.1802 5.1659 5.1485
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream d c b a 300 600 900 1200 1500 SE +/- 17.38, N = 3 SE +/- 13.91, N = 3 SE +/- 14.60, N = 4 SE +/- 14.83, N = 3 1226.80 1231.27 1231.29 1231.20
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, DistilBERT mnli - Scenario: Asynchronous Multi-Stream d c b a 12 24 36 48 60 SE +/- 0.70, N = 3 SE +/- 0.62, N = 3 SE +/- 0.60, N = 4 SE +/- 0.63, N = 3 52.07 51.92 51.93 51.93
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, DistilBERT mnli - Scenario: Synchronous Single-Stream d c b a 40 80 120 160 200 SE +/- 1.69, N = 8 SE +/- 1.76, N = 12 SE +/- 2.05, N = 12 SE +/- 1.45, N = 12 196.79 199.33 199.34 199.69
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Text Classification, DistilBERT mnli - Scenario: Synchronous Single-Stream d c b a 1.1432 2.2864 3.4296 4.5728 5.716 SE +/- 0.0457, N = 8 SE +/- 0.0484, N = 12 SE +/- 0.0573, N = 12 SE +/- 0.0391, N = 12 5.0808 5.0179 5.0196 5.0073
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream d c b a 40 80 120 160 200 SE +/- 1.48, N = 9 SE +/- 1.83, N = 3 SE +/- 2.09, N = 3 SE +/- 1.57, N = 8 181.20 177.15 180.32 183.50
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Asynchronous Multi-Stream d c b a 80 160 240 320 400 SE +/- 2.84, N = 9 SE +/- 3.47, N = 3 SE +/- 4.19, N = 3 SE +/- 3.00, N = 8 352.22 359.89 353.69 347.66
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Synchronous Single-Stream d c b a 8 16 24 32 40 SE +/- 0.23, N = 12 SE +/- 0.35, N = 6 SE +/- 0.34, N = 6 SE +/- 0.33, N = 7 35.00 35.01 35.11 34.96
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: CV Segmentation, 90% Pruned YOLACT Pruned - Scenario: Synchronous Single-Stream d c b a 7 14 21 28 35 SE +/- 0.20, N = 12 SE +/- 0.29, N = 6 SE +/- 0.29, N = 6 SE +/- 0.28, N = 7 28.56 28.55 28.47 28.59
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 400 800 1200 1600 2000 SE +/- 21.54, N = 3 SE +/- 21.19, N = 3 SE +/- 26.46, N = 3 SE +/- 21.42, N = 3 1865.02 1879.84 1875.90 1880.77
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering, Sparse INT8 - Scenario: Asynchronous Multi-Stream d c b a 8 16 24 32 40 SE +/- 0.40, N = 3 SE +/- 0.38, N = 3 SE +/- 0.48, N = 3 SE +/- 0.38, N = 3 34.25 33.99 34.08 33.98
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 20 40 60 80 100 SE +/- 0.72, N = 10 SE +/- 0.16, N = 3 SE +/- 0.30, N = 3 SE +/- 0.11, N = 3 95.70 96.55 96.38 96.43
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: BERT-Large, NLP Question Answering, Sparse INT8 - Scenario: Synchronous Single-Stream d c b a 3 6 9 12 15 SE +/- 0.08, N = 10 SE +/- 0.02, N = 3 SE +/- 0.03, N = 3 SE +/- 0.01, N = 3 10.45 10.35 10.37 10.36
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream d c b a 30 60 90 120 150 SE +/- 0.26, N = 3 SE +/- 0.19, N = 3 SE +/- 0.14, N = 3 SE +/- 0.13, N = 3 133.60 133.60 133.82 133.60
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Asynchronous Multi-Stream d c b a 100 200 300 400 500 SE +/- 0.75, N = 3 SE +/- 0.55, N = 3 SE +/- 0.39, N = 3 SE +/- 0.32, N = 3 475.45 475.35 474.37 475.13
OpenBenchmarking.org items/sec, More Is Better Neural Magic DeepSparse 1.6 Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Synchronous Single-Stream d c b a 8 16 24 32 40 SE +/- 0.33, N = 15 SE +/- 0.33, N = 15 SE +/- 0.25, N = 15 SE +/- 0.31, N = 15 33.05 31.72 31.53 31.70
OpenBenchmarking.org ms/batch, Fewer Is Better Neural Magic DeepSparse 1.6 Model: NLP Token Classification, BERT base uncased conll2003 - Scenario: Synchronous Single-Stream d c b a 7 14 21 28 35 SE +/- 0.31, N = 15 SE +/- 0.32, N = 15 SE +/- 0.25, N = 15 SE +/- 0.31, N = 15 30.29 31.56 31.73 31.58
NWChem NWChem is an open-source high performance computational chemistry package. Per NWChem's documentation, "NWChem aims to provide its users with computational chemistry tools that are scalable both in their ability to treat large scientific computational chemistry problems efficiently, and in their use of available parallel computing resources from high-performance parallel supercomputers to conventional workstation clusters." Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Seconds, Fewer Is Better NWChem 7.0.2 Input: C240 Buckyball d c b a 400 800 1200 1600 2000 1748.0 1757.3 1730.7 1744.0 1. (F9X) gfortran options: -lnwctask -lccsd -lmcscf -lselci -lmp2 -lmoints -lstepper -ldriver -loptim -lnwdft -lgradients -lcphf -lesp -lddscf -ldangchang -lguess -lhessian -lvib -lnwcutil -lrimp2 -lproperty -lsolvation -lnwints -lprepar -lnwmd -lnwpw -lofpw -lpaw -lpspw -lband -lnwpwlib -lcafe -lspace -lanalyze -lqhop -lpfft -ldplot -ldrdy -lvscf -lqmmm -lqmd -letrans -ltce -lbq -lmm -lcons -lperfm -ldntmc -lccca -ldimqm -lga -larmci -lpeigs -l64to32 -lopenblas -lpthread -lrt -llapack -lnwcblas -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz -lcomex -m64 -ffast-math -std=legacy -fdefault-integer-8 -finline-functions -O2
WRF WRF, the Weather Research and Forecasting Model, is a "next-generation mesoscale numerical weather prediction system designed for both atmospheric research and operational forecasting applications. It features two dynamical cores, a data assimilation system, and a software architecture supporting parallel computation and system extensibility." Learn more via the OpenBenchmarking.org test page.
OpenBenchmarking.org Seconds, Fewer Is Better WRF 4.2.2 Input: conus 2.5km d c b a 1200 2400 3600 4800 6000 5617.20 5583.11 5600.98 5566.73 1. (F9X) gfortran options: -O2 -ftree-vectorize -funroll-loops -ffree-form -fconvert=big-endian -frecord-marker=4 -fallow-invalid-boz -lesmf_time -lwrfio_nf -lnetcdff -lnetcdf -lmpi_usempif08 -lmpi_mpifh -lmpi -lopen-rte -lopen-pal -lhwloc -levent_core -levent_pthreads -lm -lz
a Kernel Notes: Transparent Huge Pages: madviseProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 12 December 2023 00:07 by user phoronix.
b Kernel Notes: Transparent Huge Pages: madviseProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 12 December 2023 02:28 by user phoronix.
c Kernel Notes: Transparent Huge Pages: madviseProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 12 December 2023 10:31 by user phoronix.
d Processor: 2 x INTEL XEON PLATINUM 8592+ @ 3.90GHz (128 Cores / 256 Threads), Motherboard: Quanta Cloud S6Q-MB-MPS (3B05.TEL4P1 BIOS), Chipset: Intel Device 1bce, Memory: 1008GB, Disk: 3201GB Micron_7450_MTFDKCB3T2TFS, Graphics: ASPEED, Network: 2 x Intel X710 for 10GBASE-T
OS: Ubuntu 23.10, Kernel: 6.5.0-13-generic (x86_64), Compiler: GCC 13.2.0, File-System: ext4, Screen Resolution: 1920x1080
Kernel Notes: Transparent Huge Pages: madviseCompiler 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,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-defaulted --enable-offload-targets=nvptx-none=/build/gcc-13-XYspKM/gcc-13-13.2.0/debian/tmp-nvptx/usr,amdgcn-amdhsa=/build/gcc-13-XYspKM/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-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 -vProcessor Notes: Scaling Governor: intel_pstate powersave (EPP: balance_performance) - CPU Microcode: 0x21000161Python Notes: Python 3.11.6Security Notes: gather_data_sampling: Not affected + itlb_multihit: Not affected + l1tf: Not affected + mds: Not affected + meltdown: Not affected + mmio_stale_data: 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 RSB filling PBRSB-eIBRS: SW sequence + srbds: Not affected + tsx_async_abort: Not affected
Testing initiated at 13 December 2023 00:17 by user phoronix.