1 | ID | Submitter | System | Nodes | Processor | p# | Accelerator | a# | Software | Result | Details | Code | Notes | ||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2 | Task | Image classification | Object detection | Medical imaging | Speech-to-text | Natural Language Processing | Recommendation | Large Language Model | |||||||||||||||||||||||||
3 | Data | ImageNet | OpenImages (800x800) | KiTS19 | LibriSpeech | SQuAD v1.1 | Criteo 4TB | CNN-DailyMail News | |||||||||||||||||||||||||
4 | Model | ResNet | Retinanet | 3D-UNet | RNN-T | BERT | dlrm-v2-99 | dlrm-v2-99.9 | gptj-99 | gptj-99.9 | |||||||||||||||||||||||
5 | Accuracy | 99.00 | 99.00 | 99.00 | 99.90 | 99.00 | 99.00 | 99.90 | 99.00 | 99.90 | 99.00 | 99.90 | |||||||||||||||||||||
6 | Scenario | Server | Offline | Server | Offline | Offline | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | ||||||||||||
7 | Units | Queries/s | Samples/s | Queries/s | Samples/s | Samples/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | ||||||||||||
8 | Available | ||||||||||||||||||||||||||||||||
9 | 3.1-0001 | ASUSTeK | ESC4000A-E12 (8x L4, TensorRT) | 1 | AMD EPYC 9654 96-Core Processor | 1 | NVIDIA L4 | 8 | TensorRT 9.0.0, CUDA 12.2 | 105,512.00 | 107,899.00 | 1,601.45 | 1,822.58 | 8.77 | 8.77 | 30,512.70 | 32,393.90 | 7,304.64 | 7,559.83 | 5,005.01 | 5,186.72 | details | code | ||||||||||
10 | 3.1-0002 | ASUSTeK | ESC8000A-E12 (8xH100-PCIe-80GB, TensorRT | 1 | AMD EPYC 9654 96-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 400,094.00 | 451,941.00 | 8,402.57 | 9,186.07 | 37.73 | 37.73 | 120,017.00 | 144,446.00 | 36,812.70 | 46,010.90 | 32,012.90 | 38,875.00 | 175,017.00 | 201,716.00 | 175,017.00 | 199,526.00 | details | code | ||||||
11 | 3.1-0003 | Azure | ND_H100_v5 (8x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480C | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 84.22 | 105.76 | 84.22 | 105.76 | details | code | ||||||||||||||||||
12 | 3.1-0004 | CTuning | Google Cloud Platform (g2.standard.4) | 1 | Intel(R) Xeon(R) CPU @ 2.20GHz | 1 | NVIDIA L4 | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 11,703.80 | 12,557.80 | 144.90 | 169.04 | 1.05 | 1.05 | 3,654.74 | 3,818.19 | 863.86 | 893.47 | 369.71 | 406.71 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||
13 | 3.1-0005 | CTuning | AWS cloud instance g4dn.xlarge | 1 | Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz | 1 | NVIDIA T4 | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 4,604.05 | 6,049.73 | 59.99 | 84.13 | 0.46 | 0.46 | 289.68 | 435.81 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||
14 | 3.1-0058 | Dell | Dell PowerEdge R750xa (4x A100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz | 2 | NVIDIA A100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 147,251.00 | 158,285.00 | 11,703.80 | 12,652.90 | 5,854.85 | 6,533.00 | 50,017.40 | 63,639.60 | 50,017.40 | 63,639.60 | details | code | ||||||||||||
15 | 3.1-0059 | Dell | Dell PowerEdge Server R760 (1x Intel Xeon Platinum 8480+) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 16,006.70 | 20,562.60 | 199.74 | 282.63 | 1.67 | 1,054.34 | 1,386.38 | 0.59 | 2.07 | details | code | |||||||||||||||
16 | 3.1-0060 | Dell | Dell PowerEdge Server R760 (1x Intel Xeon Platinum 8480+) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 4,204.80 | 5,784.79 | details | code | ||||||||||||||||||||||
17 | 3.1-0061 | Dell | Dell PowerEdge R750xa (4x A100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz | 2 | NVIDIA A100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 2,703.40 | 2,970.43 | 14.22 | 14.22 | 49,616.60 | 54,070.00 | 13.81 | 14.70 | details | code | ||||||||||||||
18 | 3.1-0062 | Dell | Dell PowerEdge R750xa (4x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz | 2 | NVIDIA H100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 196,555.00 | 191,776.00 | 4,280.81 | 2,035.91 | 17.81 | 17.82 | 64,620.20 | 68,571.90 | 15,321.80 | 21,392.50 | 14,966.90 | 17,895.50 | 92,240.40 | 92,569.80 | 92,240.40 | 92,569.80 | details | code | NVIDIA H100-PCIe-80GB (TDP: 310W) | |||||
19 | 3.1-0063 | Dell | Dell PowerEdge R760xa (2x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-PCIe-80GB | 2 | TensorRT 9.0.0, CUDA 12.2 | 106,002.00 | 115,122.00 | 2,248.32 | 2,308.93 | 9.40 | 9.40 | 34,054.40 | 36,405.50 | 9,146.73 | 11,834.20 | 8,252.69 | 9,824.04 | details | code | ||||||||||
20 | 3.1-0064 | Dell | Dell PowerEdge R760xa (4x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 206,018.00 | 200,132.00 | 4,354.53 | 2,557.39 | 18.66 | 18.66 | 68,021.40 | 71,644.90 | 18,069.60 | 22,940.60 | 16,306.80 | 19,467.10 | 98,916.10 | 99,964.80 | 98,916.10 | 99,964.80 | details | code | ||||||
21 | 3.1-0065 | Dell | Dell PowerEdge R760xa (4x L40, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8460Y+ | 2 | NVIDIA L40 | 4 | TensorRT 9.0.0, CUDA 12.2 | 130,021.00 | 115,291.00 | 1,901.51 | 2,069.79 | 13.21 | 13.20 | 8,619.13 | 8,311.43 | 3,505.20 | 3,853.08 | 13.81 | 18.50 | 13.81 | 18.50 | details | code | ||||||||
22 | 3.1-0066 | Dell | Dell PowerEdge XE8640 (4x NVIDIA H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8468 | 2 | NVIDIA H100-SXM-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 311,998.00 | 353,940.00 | 6,756.77 | 6,888.28 | 25.42 | 25.42 | 96,223.30 | 95,863.90 | 29,092.20 | 36,155.70 | 25,506.80 | 31,595.10 | 135,702.00 | 174,733.00 | 135,702.00 | 174,733.00 | 41.00 | 51.80 | 41.00 | 51.80 | details | code | ||
23 | 3.1-0067 | Dell | Dell PowerEdge XE9640 (4x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-SXM-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 305,053.00 | 352,037.00 | 6,656.63 | 6,869.47 | 90,021.90 | 93,774.30 | 28,011.80 | 35,402.30 | 24,807.90 | 31,369.40 | 41.00 | 51.99 | 41.00 | 51.99 | details | code | ||||||||
24 | 3.1-0068 | Dell | Dell PowerEdge XE9680 (8x A100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA A100-SXM-80GB CTS | 8 | TensorRT 9.0.0, CUDA 12.2 | 145,021.00 | 149,673.00 | 145,021.00 | 149,673.00 | 33.17 | 42.32 | details | code | NVIDIA A100-SXM4-80GB (TDP: 500W) | |||||||||||||||
25 | 3.1-0069 | Dell | Dell PowerEdge XE9680 (8x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8470 | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 620,874.00 | 706,858.00 | 12,484.80 | 14,055.80 | 51.10 | 51.10 | 178,016.00 | 187,469.00 | 57,331.00 | 70,307.40 | 51,217.90 | 62,520.90 | 326,049.00 | 344,370.00 | 326,049.00 | 344,370.00 | 81.28 | 101.83 | 81.28 | 101.83 | details | code | ||
26 | 3.1-0076 | Fujitsu | PRIMERGY_CDI_V1 (4x A100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6430 CPU @ 2.10GHz | 2 | NVIDIA A100-PCIE-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 130,599.00 | 155,688.00 | 2,916.47 | 2,972.54 | 13.92 | 13.92 | 48,216.90 | 52,506.10 | 11,848.40 | 12,496.10 | 5,787.45 | 6,416.10 | details | code | GPUs are installed in an external PCIe box. | |||||||||
27 | 3.1-0077 | GigaComputing | GIGABYTE G593-SD0 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 584,197.00 | 706,531.00 | 12,884.60 | 14,170.40 | 51.42 | 51.50 | 180,017.00 | 161,453.00 | 57,222.00 | 71,212.80 | 49,617.50 | 62,556.40 | 323,049.00 | 340,121.00 | 323,049.00 | 340,121.00 | 82.26 | 103.45 | 82.26 | 103.45 | details | code | ||
28 | 3.1-0078 | H3C | H3C UniServer R5300 G6 (8x L40, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8458P | 2 | NVIDIA L40 | 8 | TensorRT 9.0.0, CUDA 12.2 | 282,029.00 | 259,974.00 | 4,904.60 | 5,268.21 | 26.91 | 26.91 | 81,015.50 | 84,025.20 | 18,567.70 | 17,448.40 | details | code | ||||||||||||
29 | 3.1-0079 | H3C | H3C UniServer R5350 G6 (8x L40, TensorRT) | 1 | AMD EPYC 9754 128-Core Processor | 2 | NVIDIA L40 | 8 | TensorRT 9.0.0, CUDA 12.2 | 280,029.00 | 246,554.00 | 4,874.86 | 5,272.14 | 26.78 | 26.78 | 80,815.20 | 86,358.50 | 18,507.70 | 15,360.80 | details | code | ||||||||||||
30 | 3.1-0081 | HPE | 1-node-2S-SPR-PyTorch-INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 16,306.80 | 20,631.70 | 199.74 | 276.25 | 1.59 | 1,104.65 | 1,469.64 | 0.59 | 2.14 | details | code | HPE ProLiant DL380a Gen11. N/A | ||||||||||||||
31 | 3.1-0082 | HPE | 1-node-2S-SPR-PyTorch-MIX | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 4,104.84 | 5,968.03 | details | code | HPE ProLiant DL380a Gen11. N/A | |||||||||||||||||||||
32 | 3.1-0083 | HPE | HPE ProLiant DL320 Gen11 (4x L4-PCIe-24GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 5412U | 1 | NVIDIA L4-PCIe-24GB | 4 | TensorRT 9.0.0, CUDA 12.0 | 47,618.20 | 50,116.00 | 799.34 | 880.28 | 4.27 | details | code | |||||||||||||||||
33 | 3.1-0084 | HPE | HPE ProLiant DL380a Gen11 (4x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 188,015.00 | 225,970.00 | 4,004.73 | 4,516.89 | 18.85 | 18.85 | 14,007.20 | 23,441.10 | 14,007.20 | 20,329.90 | details | code | ||||||||||||
34 | 3.1-0085 | HPE | HPE ProLiant XL675d Gen10 Plus (8x A100-SXM-80GB, TensorRT) | 1 | AMD EPYC 7763 64-Core Processor | 2 | NVIDIA A100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 305,033.00 | 340,422.00 | 5,603.34 | 6,543.46 | 30.48 | 30.48 | 25,406.20 | 28,464.20 | 12,824.10 | 14,688.20 | details | code | ||||||||||||
35 | 3.1-0086 | HPE | HPE ProLiant DL385 Gen10 Plus v2 (8x QAIC100 Standard) | 1 | AMD EPYC 7543 32-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Standard | 8 | QUALCOMM Cloud AI SDK v1.9.1 | 156,018.00 | 159,270.00 | 2,229.07 | 2,277.69 | 5,479.20 | 5,917.47 | 2,728.70 | 2,956.58 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||||||||||
36 | 3.1-0088 | IEI | NF5468M6 (8x A40, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8380 | 2 | NVIDIA A40 | 8 | TensorRT 9.0.0, CUDA 12.2 | 145,820.00 | 156,472.00 | 2,453.00 | 2,554.78 | 14.88 | 15.61 | 37,912.40 | 54,707.50 | 13,155.40 | 14,022.00 | 6,004.71 | 6,863.70 | details | code | ||||||||||
37 | 3.1-0090 | Intel | 1-node-2S-SPR-PyTorch-INT4+INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 0.95 | 1.90 | details | code | QuantaGrid D54Q-2U. N/A | |||||||||||||||||||||
38 | 3.1-0091 | Intel | 1-node-2S-SPR-PyTorch-INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 16,505.90 | 20,565.50 | 214.58 | 284.75 | 1.72 | 1.72 | 1,089.77 | 1,357.33 | 4,704.13 | 5,367.77 | 0.59 | 2.05 | details | code | QuantaGrid D54Q-2U. N/A | |||||||||||
39 | 3.1-0092 | Intel | 1-node-2S-SPR-PyTorch-MIX | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | 4,204.80 | 5,782.18 | details | code | QuantaGrid D54Q-2U | |||||||||||||||||||||
40 | 3.1-0093 | Intel | 1-node-2S-SPRHBM-PyTorch-BF16 | 1 | Intel (R) Xeon (R) CPU Max 9480 | 2 | PyTorch | 0.30 | 1.03 | 0.30 | 1.03 | details | code | SC09WPRF0134SR. N/A | |||||||||||||||||||
41 | 3.1-0094 | Intel-HabanaLabs | HLS-Gaudi2-PT | 1 | Intel(R) Xeon(R) Platinum 8380 | 2 | Habana Gaudi2 | 8 | PyTorch 2.0.1a0 | 78.58 | 84.08 | 78.58 | 84.08 | details | code | ||||||||||||||||||
42 | 3.1-0095 | Krai | Dell Precision 7920 Tower (2x NVIDIA RTX A5000 GPU) | 1 | Intel(R) Xeon(R) Gold 6240R CPU @ 2.40GHz | 1 | NVIDIA RTX A5000 GPU | 2 | KRAI Inference Library Technology (KILT) with TensorRT support | 420.65 | 442.68 | 1,851.15 | 2,493.04 | 900.56 | 1,183.87 | details | code | Powered by the KRAI X and KILT technologies | |||||||||||||||
43 | 3.1-0102 | Lenovo | Lenovo ThinkSystem SR675 V3 (8x H100-PCIe-80GB, TensorRT) | 1 | AMD EPYC 9554 64-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 376,074.00 | 450,530.00 | 8,801.85 | 9,176.37 | 37.43 | 37.43 | 129,622.00 | 135,639.00 | 35,213.10 | 46,720.20 | details | code | ||||||||||||
44 | 3.1-0103 | Lenovo | Lenovo ThinkSystem SR665v1 (5x QAIC100 Pro) | 1 | AMD EPYC 75F3 32-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 5 | QUALCOMM Cloud AI SDK v1.9.1 | 115,019.00 | 116,773.00 | 1,386.48 | 1,459.08 | 3,404.83 | 3,833.70 | 1,666.36 | 1,894.73 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||||||||||
45 | 3.1-0106 | NVIDIA | ASROCKRACK 1U1G-MILAN (1x L4, TensorRT) | 1 | AMD EPYC 7313P 16-Core Processor | 2 | NVIDIA L4 | 1 | TensorRT 9.0.0, CUDA 12.2 | 12,204.40 | 12,881.70 | 199.74 | 225.92 | 1.07 | 1.07 | 3,754.56 | 3,899.48 | 898.95 | 1,028.95 | 539.24 | 631.46 | 3,305.38 | 3,672.79 | 3,305.38 | 3,672.79 | 0.89 | 1.30 | 0.89 | 1.30 | details | code | ||
46 | 3.1-0107 | NVIDIA | NVIDIA DGX H100 (1x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480C | 2 | NVIDIA H100-SXM-80GB | 1 | TensorRT 9.0.0, CUDA 12.2 | 73,019.20 | 88,526.00 | 1,621.29 | 1,728.03 | 6.45 | 6.45 | 21,510.70 | 23,306.60 | 7,003.98 | 9,102.26 | 6,104.67 | 7,877.73 | 41,516.80 | 42,856.40 | 41,516.80 | 42,856.40 | 10.15 | 13.07 | 10.15 | 13.07 | details | code | ||
47 | 3.1-0108 | NVIDIA | NVIDIA DGX H100 (8x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480C | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 584,197.00 | 704,412.00 | 12,884.60 | 14,091.40 | 51.61 | 51.61 | 144,022.00 | 151,663.00 | 56,022.10 | 70,169.70 | 49,617.50 | 62,136.70 | 315,044.00 | 329,529.00 | 315,044.00 | 329,529.00 | 82.26 | 106.32 | 82.26 | 106.32 | details | code | ||
48 | 3.1-0109 | NVIDIA | NVIDIA DGX H100 (8x H100-SXM-80GB, MaxQ, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480C | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 400,094.00 | 474,849.00 | 8,801.85 | 10,113.70 | 38.29 | 38.29 | 112,015.00 | 125,479.00 | 42,416.10 | 54,050.30 | 39,214.70 | 51,006.90 | 244,023.00 | 273,527.00 | 244,023.00 | 273,527.00 | 48.98 | 64.51 | 48.98 | 64.51 | details | code | ||
49 | 3.1-0110 | NVIDIA | NVIDIA GH200-GraceHopper-Superchip (1x GH200-96GB_aarch64, TensorRT) | 1 | NVIDIA Grace CPU | 1 | NVIDIA GH200-GraceHopper-Superchip | 1 | TensorRT 9.0.0, CUDA 12.2 | 77,018.20 | 93,198.30 | 1,731.49 | 1,849.39 | 6.76 | 6.76 | 24,008.00 | 25,974.70 | 7,704.01 | 10,163.40 | 7,003.98 | 8,645.74 | 48,516.90 | 49,001.80 | 48,516.90 | 49,001.80 | 10.96 | 13.34 | 10.96 | 13.34 | details | code | NVIDIA MGX Reference Platform | |
50 | 3.1-0111 | NVIDIA | Gigabyte G482-Z54 (1x H100-PCIe-80GB, TensorRT) | 1 | AMD EPYC 7742 64-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 1 | TensorRT 9.0.0, CUDA 12.2 | 47,017.50 | 54,851.40 | 1,049.36 | 1,111.11 | 4.55 | 4.55 | 15,006.80 | 17,106.70 | 4,564.11 | 5,711.00 | 4,004.73 | 4,961.92 | 24,507.70 | 25,153.40 | 24,507.70 | 25,153.40 | details | code | ||||||
51 | 3.1-0112 | NVIDIA | Gigabyte G482-Z54 (8x H100-PCIe-80GB, TensorRT) | 1 | AMD EPYC 7742 64-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 368,074.00 | 442,960.00 | 8,402.57 | 9,176.43 | 37.02 | 37.02 | 100,015.00 | 115,297.00 | 35,373.40 | 45,698.50 | 32,012.90 | 39,412.80 | 170,013.00 | 192,829.00 | 170,013.00 | 192,829.00 | details | code | ||||||
52 | 3.1-0113 | NVIDIA | Gigabyte G482-Z54 (8x H100-PCIe-80GB, MaxQ, TensorRT) | 1 | AMD EPYC 7742 64-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 240,024.00 | 348,572.00 | 6,303.18 | 6,719.02 | 27.36 | 27.36 | 88,023.00 | 98,684.60 | 33,011.50 | 39,646.90 | 28,513.00 | 34,374.10 | 132,024.00 | 162,281.00 | 132,024.00 | 162,281.00 | 40.01 | 50.57 | 40.01 | 50.57 | details | code | ||
53 | 3.1-0118 | Nutanix | NX_3155G_G8_A100_PCIe_80GBx2 | 1 | Intel(R) Xeon(R) Gold 6354 CPU @ 3.00GHz | 2 | NVIDIA A100-PCIe-80GB | 2 | TensorRT 8.6.0, CUDA 12.0 | 64,520.40 | 74,652.60 | 1,250.31 | 1,297.28 | 6.94 | 6.94 | 24,008.00 | 26,250.10 | 5,603.34 | 6,241.74 | 2,803.89 | 3,275.44 | details | code | ||||||||||
54 | 3.1-0119 | Oracle | BM.GPU.A10.4 | 1 | Intel(R) Xeon(R) Platinum 8358 CPU @ 2.60GHz | 2 | NVIDIA A10-PCI-24GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 855.00 | 953.53 | 5.15 | 9,202.52 | 16,989.30 | details | code | |||||||||||||||||
55 | 3.1-0120 | Oracle | BM.GPU.A100-v2.8 | 1 | AMD EPYC 7J13 64-Core Processor | 2 | NVIDIA A100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 290,028.00 | 325,567.00 | 5,603.34 | 6,512.98 | 30.31 | 30.33 | 104,012.00 | 107,408.00 | 25,406.20 | 28,028.60 | 12,824.10 | 14,534.40 | 80,018.10 | 138,331.00 | 80,018.10 | 138,179.00 | 16.92 | 27.13 | 17.04 | 25.29 | details | code | ||
56 | 3.1-0121 | Oracle | BM.GPU.H100.8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 584,197.00 | 703,548.00 | 12,884.60 | 14,047.20 | 51.45 | 51.48 | 56,022.10 | 70,689.90 | 49,617.50 | 62,285.50 | 300,033.00 | 339,265.00 | 300,033.00 | 339,050.00 | 79.90 | 106.69 | details | code | ||||||
57 | 3.1-0122 | Qualcomm | GIGABYTE G292-Z43 (16x QAIC100 Pro) | 1 | AMD EPYC 7713 64-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 16 | QUALCOMM Cloud AI SDK v1.9.1 | 370,071.00 | 398,010.00 | 4,578.88 | 4,671.42 | 12,003.70 | 12,536.80 | 5,934.41 | 6,315.87 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||||||||||
58 | 3.1-0123 | Qualcomm | GIGABYTE G292-Z43 (16x QAIC100 Pro, EE) | 1 | AMD EPYC 7742 64-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 16 | QUALCOMM Cloud AI SDK v1.9.1 | 328,050.00 | 337,737.00 | 3,804.75 | 3,949.99 | 9,777.31 | 10,068.20 | 5,354.05 | 5,584.38 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||||||||||
59 | 3.1-0124 | Qualcomm | GIGABYTE R282-Z93 (8x QAIC100 Pro, EE) | 1 | AMD EPYC 7282 16-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 8 | QUALCOMM Cloud AI SDK v1.9.1 | 148,018.00 | 169,969.00 | 1,901.51 | 1,975.93 | 4,804.25 | 5,031.42 | 2,778.99 | 2,915.72 | details | code | With 75W Accelerator TDP constraints. 3x QAIC100 on riser CRS2033; 3x QAIC100 on riser CRS2033; 2x QAIC100 on riser CRS2026. Powered by the KRAI X and KILT technologies | |||||||||||||
60 | 3.1-0128 | Quanta_Cloud_Technology | 1-node-2S-SPR-PyTorch-INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ 56-Core Processor | 2 | PyTorch | 16,505.90 | 20,282.00 | 211.65 | 288.02 | 4,104.84 | 5,643.03 | 1,079.75 | 1,354.06 | 4,504.80 | 4,767.43 | 0.59 | 2.07 | details | code | QuantaGrid D54Q-2U | |||||||||||
61 | 3.1-0129 | Quanta_Cloud_Technology | D54Q_2U (2x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6430 32-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 2 | TensorRT 9.0.0, CUDA 12.2 | 94,018.40 | 112,561.00 | 2,001.26 | 2,265.11 | 9.34 | 9.34 | 30,013.00 | 36,083.80 | 9,002.36 | 11,251.90 | 8,003.55 | 9,491.09 | details | code | ||||||||||
62 | 3.1-0130 | Quanta_Cloud_Technology | D54Q_2U (4x L4-PCIe-24GB, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6430 32-Core Processor | 2 | NVIDIA L4-PCIe-24GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 48,818.30 | 51,005.20 | 799.34 | 889.04 | 4.32 | 4.32 | 15,006.80 | 15,709.20 | 3,604.96 | 3,733.81 | 2,162.57 | 2,554.79 | 3.44 | 5.21 | details | code | ||||||||
63 | 3.1-0132 | Supermicro | AS-8125GS-TNHR (8x H100-SXM-80GB, TensorRT) | 1 | AMD EPYC 9554 | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 590,083.00 | 707,537.00 | 12,996.70 | 14,136.20 | 51.85 | 51.85 | 172,014.00 | 176,599.00 | 56,893.70 | 70,619.70 | 50,617.40 | 62,456.10 | 322,447.00 | 342,065.00 | 325,049.00 | 341,806.00 | 84.50 | 105.53 | 84.50 | 105.91 | details | code | ||
64 | 3.1-0133 | Supermicro | SYS-421GU-TNXR (4x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8470Q | 2 | NVIDIA H100-SXM-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 40.01 | 37.93 | 40.01 | 37.93 | details | code | ||||||||||||||||||
65 | 3.1-0134 | Supermicro | SYS-521GE-TNRT (8xH100-PCIe-80GB) | 1 | Intel(R) Xeon(R) Platinum 8462Y+ | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 368,074.00 | 446,436.00 | 8,402.57 | 9,170.89 | 37.64 | 37.64 | 100,015.00 | 131,664.00 | 35,012.80 | 46,250.70 | 30,512.70 | 40,132.90 | 170,013.00 | 198,707.00 | 170,013.00 | 198,707.00 | details | code | ||||||
66 | 3.1-0135 | Supermicro | SYS-821GE-TNHR (8x H100-SXM-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8468 | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 593,579.00 | 705,305.00 | 13,020.70 | 14,195.80 | 51.99 | 51.98 | 57,101.70 | 70,682.60 | 50,969.10 | 62,479.80 | 327,051.00 | 340,928.00 | 324,250.00 | 340,658.00 | 85.57 | 107.33 | 85.43 | 107.06 | details | code | ||||
67 | 3.1-0136 | TTA | KR580S1 | 1 | Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz | 2 | NVIDIA T4 | 2 | TensorRT 8.6.0, CUDA 12.0 | 8,202.98 | 10,300.70 | details | code | ||||||||||||||||||||
68 | 3.1-0137 | TTA | KR580S1 | 1 | Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz | 2 | NVIDIA T4 | 2 | tensorrt 8.6.1, | 8,701.87 | 10,639.80 | 130.09 | 160.77 | 0.87 | 539.24 | 742.60 | 249.41 | 359.26 | details | code | Powered by MLCommons CM automation language and CK playground | ||||||||||||
69 | 3.1-0138 | xFusion | xFusion FusionServer G5500V7(10x NVIDIA A30, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6458Q @ 3.1 GHz | 2 | NVIDIA A30 | 10 | TensorRT 9.0.0, CUDA 12.2 | 184,417.00 | 194,705.00 | 3,864.53 | 3,897.79 | 17.92 | 17.92 | 58,821.10 | 72,693.30 | 15,407.70 | 17,263.50 | 6,723.79 | 8,687.66 | 53,519.80 | 68,727.90 | 53,519.80 | 68,727.90 | details | code | ||||||
70 | 3.1-0139 | xFusion | xFusion FusionServer G5500V7(8x NVIDIA A30, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6458Q @ 3.1 GHz | 2 | NVIDIA A30 | 8 | TensorRT 9.0.0, CUDA 12.2 | 147,668.00 | 155,998.00 | 3,100.41 | 3,047.61 | 14.32 | 14.32 | 46,517.00 | 58,061.90 | 12,304.90 | 13,609.60 | 5,504.20 | 6,954.59 | 42,516.50 | 55,253.20 | 42,516.50 | 55,253.20 | 8.93 | 9.83 | 8.93 | 9.82 | details | code | ||
71 | 3.1-0140 | xFusion | xFusion FusionServer G5500V7(10x NVIDIA L40, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6458Q @ 3.1 GHz | 2 | NVIDIA L40 | 10 | TensorRT 9.0.0, CUDA 12.2 | 295,531.00 | 258,506.00 | 5,264.69 | 4,991.75 | 28.65 | 28.87 | 94,219.30 | 97,499.00 | 20,508.90 | 17,996.30 | 8,902.43 | 8,527.61 | 66,022.20 | 82,179.50 | 66,022.20 | 82,179.50 | details | code | ||||||
72 | 3.1-0141 | xFusion | xFusion FusionServer G5500V7(8x NVIDIA L40, TensorRT) | 1 | Intel(R) Xeon(R) Gold 6458Q @ 3.1 GHz | 2 | NVIDIA L40 | 8 | TensorRT 9.0.0, CUDA 12.2 | 234,743.00 | 206,778.00 | 4,204.80 | 3,981.97 | 23.07 | 23.07 | 75,540.40 | 83,390.70 | 16,726.80 | 14,439.70 | 9,701.77 | 8,797.10 | 64,021.20 | 67,717.70 | 64,021.20 | 67,717.70 | 30.15 | 40.19 | 30.15 | 40.19 | details | code | ||
73 | 3.1-0142 | xFusion | xFusion FusionServer 2288H V7(6x NVIDIA L4, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8458P CPU @ 2.7 GHz | 2 | NVIDIA L4 | 6 | TensorRT 9.0.0, CUDA 12.2 | 74,520.70 | 76,663.90 | 1,280.38 | 1,227.30 | 6.82 | 6.82 | 14,807.60 | 24,375.60 | 5,404.18 | 5,748.72 | 3,754.56 | 3,942.95 | 8,003.55 | 8,914.83 | 8,003.55 | 8,903.54 | 6.95 | 8.83 | 6.95 | 8.85 | details | code | ||
74 | Preview | ||||||||||||||||||||||||||||||||
75 | 3.1-0143 | tpu-v5e-4 | 1 | AMD EPYC 7B13 | 1 | TPU v5e | 4 | SAX | 7.13 | 9.81 | details | code | |||||||||||||||||||||
76 | 3.1-0144 | Quanta_Cloud_Technology | D54U-3U (4x H100-PCIe-80GB, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | NVIDIA H100-PCIe-80GB | 4 | TensorRT 9.0.0, CUDA 12.2 | 188,015.00 | 221,741.00 | 4,004.73 | 4,503.27 | 18.31 | 18.31 | 60,023.10 | 70,529.20 | 18,006.50 | 22,801.60 | 16,006.70 | 19,285.10 | details | code |
1 | ID | Submitter | System | Nodes | Processor | p# | Accelerator | a# | Software | Results | Details | Code | Notes | ||||||||||||||||||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2 | Task | Image classification | Object detection | Medical imaging | Speech-to-text | Natural Language Processing | Recommendation | Large Language Model | |||||||||||||||||||||||||||||||||||||||||||||
3 | Data | ImageNet | OpenImages (800x800) | KiTS19 | LibriSpeech | SQuAD v1.1 | Criteo 4TB | CNN-DailyMail News | |||||||||||||||||||||||||||||||||||||||||||||
4 | Model | ResNet | Retinanet | 3D-UNet | RNN-T | BERT | dlrm-v2-99 | gptj-99 | |||||||||||||||||||||||||||||||||||||||||||||
5 | Accuracy (%FP32 ref) | 99.00 | 99.00 | 99.00 | 99.90 | 99.00 | 99.00 | 99.90 | 99.00 | 99.90 | gptj-99 | gptj-99.9 | |||||||||||||||||||||||||||||||||||||||||
6 | Scenario | Server | Offline | Server | Offline | Offline | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | ||||||||||||||||||||||||||||||||
7 | Units | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | samples/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | ||||||||||||
8 | Available | ||||||||||||||||||||||||||||||||||||||||||||||||||||
9 | 3.1-0109 | NVIDIA | NVIDIA DGX H100 (8x H100-SXM-80GB, MaxQ, TensorRT) | 1 | Intel(R) Xeon(R) Platinum 8480C | 2 | NVIDIA H100-SXM-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 400,094.00 | 4,135.72 | 474,849.00 | 4,063.98 | 8,801.85 | 4,545.92 | 10,113.70 | 4,554.48 | 38.29 | 4,166.64 | 38.29 | 4,166.64 | 112,015.00 | 4,400.42 | 125,479.00 | 4,199.93 | 42,416.10 | 5,223.90 | 54,050.30 | 5,038.23 | 39,214.70 | 5,528.30 | 51,006.90 | 5,594.39 | 244,023.00 | 5,794.91 | 273,527.00 | 5,629.87 | 244,023.00 | 5,794.91 | 273,527.00 | 5,629.87 | 48.98 | 3,830.87 | 64.51 | 3,805.44 | 48.98 | 3,830.87 | 64.51 | 3,805.44 | details | code | ||
10 | 3.1-0113 | NVIDIA | Gigabyte G482-Z54 (8x H100-PCIe-80GB, MaxQ, TensorRT) | 1 | AMD EPYC 7742 64-Core Processor | 2 | NVIDIA H100-PCIe-80GB | 8 | TensorRT 9.0.0, CUDA 12.2 | 240,024.00 | 2,272.53 | 348,572.00 | 2,268.11 | 6,303.18 | 2,347.19 | 6,719.02 | 2,254.60 | 27.36 | 2,144.75 | 27.36 | 2,144.75 | 88,023.00 | 2,248.54 | 98,684.60 | 2,235.99 | 33,011.50 | 3,348.38 | 39,646.90 | 3,047.66 | 28,513.00 | 3,116.67 | 34,374.10 | 3,310.20 | 132,024.00 | 3,049.98 | 162,281.00 | 2,955.39 | 132,024.00 | 3,049.98 | 162,281.00 | 2,955.39 | 40.01 | 2,187.31 | 50.57 | 2,195.34 | 40.01 | 2,187.31 | 50.57 | 2,195.34 | details | code | ||
11 | 3.1-0123 | Qualcomm | GIGABYTE G292-Z43 (16x QAIC100 Pro, EE) | 1 | AMD EPYC 7742 64-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 16 | QUALCOMM Cloud AI SDK v1.9.1 | 328,050.00 | 1,417.29 | 337,737.00 | 1,425.32 | 3,804.75 | 980.13 | 3,949.99 | 982.86 | 9,777.31 | 1,091.04 | 10,068.20 | 1,098.44 | 5,354.05 | 1,163.16 | 5,584.38 | 1,191.47 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||||||||||||||||||||||
12 | 3.1-0124 | Qualcomm | GIGABYTE R282-Z93 (8x QAIC100 Pro, EE) | 1 | AMD EPYC 7282 16-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 8 | QUALCOMM Cloud AI SDK v1.9.1 | 148,018.00 | 631.21 | 169,969.00 | 686.11 | 1,901.51 | 455.41 | 1,975.93 | 469.22 | 4,804.25 | 529.57 | 5,031.42 | 534.62 | 2,778.99 | 599.40 | 2,915.72 | 617.11 | details | code | With 75W Accelerator TDP constraints. 3x QAIC100 on riser CRS2033; 3x QAIC100 on riser CRS2033; 2x QAIC100 on riser CRS2026. Powered by the KRAI X and KILT technologies |
1 | ID | Submitter | System | Nodes | Processor | p# | Accelerator | a# | Software | UsedModel | Accuracy | Result | Details | Code | Notes | |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2 | Task | Image classification | Object detection | Medical imaging | Speech-to-text | Natural Language Processing | Recommendation | Large Language Model | ||||||||||||||||||||||
3 | Data | ImageNet | OpenImages (800x800) | KiTS19 | LibriSpeech | SQuAD v1.1 | Criteo 4TB | CNN-DailyMail News | ||||||||||||||||||||||
4 | Model | ResNet | Retinanet | 3D-UNet | RNN-T | BERT | dlrm-v2-99 | gptj-99 | ||||||||||||||||||||||
5 | Accuracy | 99.00 | 99.00 | 99.00 | 99.90 | 99.00 | 99.00 | 99.90 | 99.00 | 99.00 | ||||||||||||||||||||
6 | Scenario | Server | Offline | Server | Offline | Offline | Offline | Server | Offline | Server | Offline | Server | Offline | Server | Offline | Offline | ||||||||||||||
7 | Units | Queries/s | Samples/s | Queries/s | Samples/s | Samples/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | Samples/s | ||||||||||||||
8 | Available | |||||||||||||||||||||||||||||
9 | 3.1-0148 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 3d-unet-99 | 0.86 | 4.18 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||
10 | 3.1-0149 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 3d-unet-99.9 | 0.86 | 4.18 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||
11 | 3.1-0150 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | bert-99 | 90.36 | 4144.88 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||
12 | 90.51 | 3954.44 | details | code | ||||||||||||||||||||||||||
13 | 3.1-0151 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | bert-99.9 | 90.87 | 1521.60 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||
14 | 90.88 | 1682.05 | details | code | ||||||||||||||||||||||||||
15 | 3.1-0152 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | resnet50 | 76.16 | 37013.80 | 45772.10 | details | code | Powered by MLCommons CM automation language and CK playground. | ||||||||||||||
16 | 3.1-0153 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | retinanet | 37.39 | 615.99 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||
17 | 37.41 | 589.09 | details | code | ||||||||||||||||||||||||||
18 | 3.1-0154 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | rnnt | 92.55 | 14157.00 | 15377.20 | details | code | Powered by MLCommons CM automation language and CK playground. | ||||||||||||||
19 | 3.1-4184 | Dell | Dell PowerEdge Server R760 (Intel Xeon Platinum 8480+) | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | dlrm-v2-99 | 80.23 | 4404.50 | 5016.63 | details | code | |||||||||||||||||
20 | 3.1-4185 | Intel | 1-node-2S-SPR-PyTorch-INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 2 | PyTorch | bert-99.9_MiniLM | 90.80 | 6543.62 | details | code | QuantaGrid D54Q-2U. N/A | |||||||||||||||||
21 | 3.1-4187 | Moffett | H3C R5300 G5 (1x SparseOne S30, PCIe/FHFL, Moffett-SDK) | 1 | Intel(R) Xeon(R) Gold 6348 | 2 | MOFFETT S30-PCIe/FHFL-60GB | 1 | Moffett SDK | gptj-99 | 42.96 | 23.28 | details | code | ||||||||||||||||
22 | 3.1-4188 | Moffett | H3C R5300 G5 (4x SparseOne S30, PCIe/FHFL, Moffett-SDK) | 1 | Intel(R) Xeon(R) Gold 6348 | 2 | MOFFETT S30-PCIe/FHFL-60GB | 4 | Moffett SDK | gptj-99 | 42.96 | 91.57 | details | code | ||||||||||||||||
23 | 3.1-4189 | Moffett | Inspur NF5468M6 (8x SparseOne S30, PCIe/FHFL, Moffett-SDK) | 1 | Intel(R) Xeon(R) Platinum 8380 CPU @ 2.30GHz | 2 | MOFFETT S30-PCIe/FHFL-60GB | 8 | Moffett SDK | gptj-99 | 42.96 | 170.59 | details | code | ||||||||||||||||
24 | 3.1-4190 | NeuralMagic | aws.c6g_2xlarge | 1 | ARM Neoverse-N1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50-none-vnni-bert-99 | 90.43 | 32.98 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
25 | 3.1-4191 | NeuralMagic | aws.c6g_2xlarge | 1 | ARM Neoverse-N1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50_quant-none-vnni-bert-99 | 90.35 | 72.32 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
26 | 3.1-4192 | NeuralMagic | aws.c6g_2xlarge | 1 | ARM Neoverse-N1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-base_quant-none-bert-99 | 90.78 | 39.83 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
27 | 3.1-4193 | NeuralMagic | aws.c6g_2xlarge | 1 | ARM Neoverse-N1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-none-base-none-bert-99 | 90.89 | 18.86 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
28 | 3.1-4194 | NeuralMagic | aws.c6g_2xlarge | 1 | ARM Neoverse-N1 | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned97-quant-none-bert-99 | 90.09 | 9.56 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
29 | 3.1-4195 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | bert-large-pruned80_quant-none-vnni-bert-99 | 90.23 | 10.54 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
30 | 3.1-4196 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50-none-vnni-bert-99 | 90.43 | 49.28 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
31 | 3.1-4197 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50_quant-none-vnni-bert-99 | 90.35 | 107.21 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
32 | 3.1-4198 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-base_quant-none-bert-99 | 90.78 | 56.11 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
33 | 3.1-4199 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | mobilebert-none-base-none-bert-99 | 90.89 | 21.50 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
34 | 3.1-4200 | NeuralMagic | aws.c7g_2xlarge | 1 | ARM Neoverse-V1 | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned97-quant-none-bert-99 | 90.09 | 15.91 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
35 | 3.1-4201 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | bert-base-pruned90-none-bert-99 | 88.42 | 29.85 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
36 | 3.1-4202 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | bert-base-pruned95_obs_quant-none-bert-99 | 87.89 | 77.47 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
37 | 3.1-4203 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | bert-base_cased-pruned90-none-bert-99 | 4.53 | 28.84 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
38 | 3.1-4204 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | bert-large-base-none-bert-99 | 89.65 | 3.08 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
39 | 3.1-4205 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | bert-large-pruned80_quant-none-vnni-bert-99 | 90.27 | 20.13 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
40 | 3.1-4206 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50-none-vnni-bert-99 | 90.43 | 77.57 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
41 | 3.1-4207 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | mobilebert-14layer_pruned50_quant-none-vnni-bert-99 | 90.40 | 158.57 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
42 | 3.1-4208 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | mobilebert-base_quant-none-bert-99 | 90.79 | 88.67 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
43 | 3.1-4209 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | mobilebert-none-base-none-bert-99 | 90.89 | 38.77 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
44 | 3.1-4210 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-base-pruned90-none-bert-99 | 88.31 | 29.66 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
45 | 3.1-4211 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-large-base-none-bert-99 | 89.65 | 3.08 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
46 | 3.1-4212 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned95-none-vnni-bert-99 | 90.18 | 13.08 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
47 | 3.1-4213 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned95_quant-none-vnni-bert-99 | 90.03 | 30.74 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
48 | 3.1-4214 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned97-none-bert-99 | 90.14 | 14.97 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
49 | 3.1-4215 | NeuralMagic | gcp.c3_standard_8 | 1 | Intel(R) Xeon(R) Platinum 8481C CPU @ 2.70GHz | 1 | deepsparse v1.6.0.20230801 | obert-large-pruned97-quant-none-bert-99 | 90.18 | 27.09 | details | code | Powered by MLCommons Collective Mind framework (CK2). | |||||||||||||||||
50 | 3.1-4240 | Supermicro | 1-node-4S-SPR-PyTorch-INT8 | 1 | Intel(R) Xeon(R) Platinum 8480+ | 4 | PyTorch | gptj-99 | 42.92 | 2.81 | details | code | ||||||||||||||||||
51 | RDI | |||||||||||||||||||||||||||||
52 | 3.1-4242 | NVIDIA | NVIDIA L4 (1x L4, TensorRT) | 1 | AMD EPYC 7313P 16-Core Processor | 2 | NVIDIA L4 | 1 | TensorRT 9.0.0, CUDA 12.2 | bert-99 | 90.17 | 4,264.85 | details | code | ||||||||||||||||
53 | 90.18 | 4,609.04 | details | code | ||||||||||||||||||||||||||
54 | ||||||||||||||||||||||||||||||
55 | ||||||||||||||||||||||||||||||
56 | ||||||||||||||||||||||||||||||
57 | ||||||||||||||||||||||||||||||
58 | ||||||||||||||||||||||||||||||
59 | ||||||||||||||||||||||||||||||
60 | ||||||||||||||||||||||||||||||
61 | ||||||||||||||||||||||||||||||
62 |
1 | ID | Submitter | System | Nodes | Processor | p# | Accelerator | a# | Software | UsedModel | Accuracy | Result | Details | Code | Notes | ||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2 | Task | Image classification | Object detection | Medical imaging | Speech-to-text | Natural Language Processing | |||||||||||||||||||||||||||||||||
3 | Data | ImageNet | OpenImages (800x800) | KiTS19 | LibriSpeech | SQuAD v1.1 | |||||||||||||||||||||||||||||||||
4 | Model | ResNet | Retinanet | 3D-UNet | RNN-T | BERT | |||||||||||||||||||||||||||||||||
5 | Accuracy (%FP32 ref) | 99.00 | 99.00 | 99.00 | 99.90 | 99.00 | 99.00 | 99.90 | |||||||||||||||||||||||||||||||
6 | Scenario | Server | Offline | Server | Offline | Offline | Offline | Server | Offline | Server | Offline | Server | Offline | ||||||||||||||||||||||||||
7 | Units | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | samples/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | queries/s | System Power (W) | samples/s | System Power (W) | ||||||||||||||
8 | Available | ||||||||||||||||||||||||||||||||||||||
9 | 3.1-0148 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 3d-unet-99 | 0.86 | 4.18 | 601.89 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||||
10 | 3.1-0149 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | 3d-unet-99.9 | 0.86 | 4.18 | 601.89 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||||
11 | 3.1-0150 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | bert-99 | 90.36 | 4144.88 | 625.26 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||||
12 | 90.51 | 3954.44 | 621.15 | details | code | ||||||||||||||||||||||||||||||||||
13 | 3.1-0151 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | bert-99.9 | 90.87 | 1521.60 | 620.65 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||||
14 | 90.88 | 1682.05 | 612.72 | details | code | ||||||||||||||||||||||||||||||||||
15 | 3.1-0152 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | resnet50 | 76.16 | 37013.80 | 581.93 | 45772.10 | 617.54 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||
16 | 3.1-0153 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | retinanet | 37.39 | 615.99 | 582.65 | details | code | Powered by MLCommons CM automation language and CK playground. | |||||||||||||||||||||||
17 | 37.41 | 589.09 | 566.96 | details | code | ||||||||||||||||||||||||||||||||||
18 | 3.1-0154 | CTuning | PCSPECIALIST AMD AM5 PC with Nvidia RTX 4090 | 1 | AMD Ryzen 9 7950X 16-Core Processor | 1 | NVIDIA GeForce RTX 4090 (Ada Lovelace) | 1 | Nvidia inference implementation with CM API, TensorRT v8.6.1.6 | rnnt | 92.55 | 14157.00 | 621.62 | 15377.20 | 615.30 | details | code | Powered by MLCommons CM automation language and CK playground. |
1 | ID | Submitter | System | Nodes | Processor | p# | Accelerator | a# | Software | Result | Details | Code | Notes | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2 | Task | Object Detection | Natural Language Processing | ||||||||||||||||
3 | Data | Open Images (800x800) | SQuAD v1.1 | ||||||||||||||||
4 | Model | Retinanet | bert-99 | bert-99.9 | |||||||||||||||
5 | Accuracy | 99.00 | 99.00 | 99.90 | |||||||||||||||
6 | Scenario | Server | Offline | Server | Offline | Server | Offline | ||||||||||||
7 | Units | Queries/s | Samples/s | Queries/s | Samples/s | Queries/s | Samples/s | ||||||||||||
8 | Available | ||||||||||||||||||
9 | 3.1-0145 | HPE | HPE ProLiant DL385 Gen10 Plus v2 (2x QAIC100 Standard) | 1 | AMD EPYC 7543 32-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Standard | 2 | QUALCOMM Cloud AI SDK v1.9.1 | 510.45 | 568.11 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||||
10 | 3.1-0146 | HPE | HPE ProLiant DL385 Gen10 Plus v2 (8x QAIC100 Standard) | 1 | AMD EPYC 7543 32-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Standard | 8 | QUALCOMM Cloud AI SDK v1.9.1 | 5,504.20 | 5,906.87 | 2,703.40 | 2,956.35 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies | |||
11 | 3.1-0147 | Qualcomm | Gigabyte G292-Z43 (16x QAIC100) | 1 | AMD EPYC 7713 64-Core Processor | 2 | QUALCOMM Cloud AI 100 PCIe/HHHL Pro | 16 | QUALCOMM Cloud AI SDK v1.9.1 | 4,583.88 | 4,679.51 | 12,003.70 | 12,541.20 | 6,004.71 | 6,313.36 | details | code | With 75W Accelerator TDP constraints. Powered by the KRAI X and KILT technologies |