

| FP64 peak performance | 9.7 TF | | | |
| FP64 Tensor Core peak performance | 19.5 TF | | | |
| FP32 peak performance | 19.5 TF | | | |
| FP32 Tensor Core peak performance | 156 TF | 312 TF* | | | |
| BFLOAT16 Tensor Core peak performance | 312 TF | 624 TF* | | | |
| FP16 Tensor Core peak performance | 312 TF | 624 TF* | | | |
| INT8 Tensor Core peak performance | 624 TOPS | 1,248 TOPS* | | | |
| INT4 Tensor Core peak performance | 1,248 TOPS | 2,496 TOPS* | | | |
| GPU memory | 40GB | | | |
| GPU memory bandwidth | 1,555 GB/s | | | |
| interconnected | NVIDIA NVLink 600 GB/s** PCIe Gen4 64 GB/s | | | |
| Multi-instance GPU | Various instance sizes with up to 7 MIGs at 5GB | | | |
| Form factor | PCIe | | | |
| Max TDP power consumption | 250 W | | | |












