| Model | GPU Architecture | Tensor Cores | CUDA Cores | DP Perf. | SP Perf. | Tensor Perf. | Integer Op. | Half-Precision Perf. | GPU Memory | Memory Bandwidth | Interconnect Bandwidth | Form Factor | Max Power Consumption |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Tesla P100 | Pascal | - | 3584 | 4.7 TFLOPS | 9.3 TFLOPS | - | - | 18.7 TeraFLOPS | 16GB HBM2/12GB HBM2 | - | - | Full Height | 250 W |
| TeslA P4 | Pascal | - | - | - | 5.5 TFLOPS | - | 22 TOPS | - | 8 GB | 192 GB/s | - | Low Profile | 50W/75W |
| Tesla P40 | Pascal | - | - | - | 12 TeraFLOPS | - | 47 TOPS | - | 24 GB | 346 GB/s | - | Full Height | 250 W |
| TESLA V100 | Volta | 640 | 5120 | 7 TFLOPS | 14 TFLOPS | 112 TFLOPS | - | - | 32GB /16GB HBM2 | 900 GB/sec | 32 GB/sec | Full Height | 250 W |