nvidia-smi
Driver Version: 595.71 | CUDA Version: 13.2
GPU NameTemp | Power | Util
0 H200 SXM 141GB62°C | 891W / 1000W | 98%
1 H200 SXM 141GB64°C | 878W / 1000W | 99%
2 H200 SXM 141GB61°C | 903W / 1000W | 97%
3 H200 SXM 141GB65°C | 867W / 1000W | 99%
4 H200 SXM 141GB63°C | 912W / 1000W | 98%
5 H200 SXM 141GB60°C | 875W / 1000W | 99%
6 H200 SXM 141GB66°C | 895W / 1000W | 97%
7 H200 SXM 141GB63°C | 882W / 1000W | 99%
VRAM GPU-0133120 MiB / 143360 MiB
VRAM GPU-1 ~ 7 (avg)130048 MiB / 143360 MiB
Processes
GPUPIDProcess NameMiB
0-731024torchrun pretrain.py131072×8
# No idle VRAM. This is how it should be.
$