Production GPU Clusters
| Zone Sort descending | Nodes | Total Memory (GB) | GPU Architecture | Total GPUs | GPUs per compute node | Cores/ Node | Total Cores | Memory/ Node | GPU peak performance (TFLOP/s double precision) | GPU global memory (GB) | Switch | Peak TFLOPs (GPUs) | Peak TFLOPS (CPUs+GPUs) | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Vertex | 40 | 12,256 | NVIDIA Tesla | 128 | 4 | 16 | 1,024 | 383 | 32 | 1 002.8 | 1 030.6 | |||
| Pascal | CZ | 164 | 41,984 | NVIDIA Tesla P100 (Pascal) | 326 | 2 | 36 | 5,904 | 256 | 5.00 | 16 | Cornelis Networks Omni-Path | 1 727.8 | 1 926.1 |
| Lassen | CZ | 795 | 253,440 | NVIDIA V100 (Volta) | 3,168 | 4 | 44 | 34,848 | 256 | 7.00 | IB EDR | 22 192.2 | 23 047.2 | |
| Tioga | CZ | 38 | 12,288 | AMD MI-250X | 152 | 4 | 64 | 2,432 | 512 | 45.00 | 128 | HPE Slingshot 11 | 6 840.0 | 6 916.0 |
| RZAnsel | RZ | 62 | 17,280 | NVIDIA V100 (Volta) | 216 | 4 | 44 | 2,376 | 256 | 7.00 | IB EDR | 1 512.0 | 1 570.0 | |
| RZWhamo | RZ | 37 | 9,472 | Nvidia V100s / AMD Mi60 | 64 | 2 | 64 | 2,368 | 256 | 24.00 | 2 | Mellanox HDR / Cray Slingshot | 761.6 | 843.5 |
| RZNevada | RZ | 24 | 6,144 | AND MI-100 | 24 | 1 | 64 | 1,536 | 256 | 12.00 | 32 | HPE Slingshot 11 | 276.0 | 324.0 |
| RZVernal | RZ | 32 | 19,456 | AMD MI-250X | 128 | 4 | 64 | 2,048 | 512 | 45.00 | 128 | HPE Slingshot 11 | 5 760.0 | 5 824.0 |
| Sierra | SCF | 4,474 | 1,382,400 | NVIDIA V100 (Volta) | 17,280 | 4 | 44 | 190,080 | 256 | 7.00 | IB EDR | 120 960.0 | 125 626.0 | |
| Tenaya | SCF | 24 | 16,384 | AMD MI-250X | 96 | 4 | 64 | 1,536 | 512 | 45.00 | 128 | HPE Slingshot 11 | 4 320.0 | 4 368.0 |