NVIDIA / ACCELERATOR VARIANT

H200.

NVIDIA H200-NVL-141GB · 141 GB · NVL

Logical ID accelerator:nvidia-h200-nvl-141-gb
Content version d8f08cb219662ecc800f46b6c79ce2db16dc1fa844d32774381dec121c214f24

Submitted systems

Cisco · X210M8_X580P_H200NVL_141GBx4_TRT

1 nodes × 4 accelerators = 4 total

Topology: 4th Gen NVLink, 600GB/s
Software: TensorRT 10.11, CUDA 12.9

Pinned system specification

llama3.1-8b · Offline: 24,216.6 tokens/s · source

llama3.1-8b · Server: 25,731.6 tokens/s · source

Dell · XE7740_H200_NVL_141GBx8_TRT

1 nodes × 8 accelerators = 8 total

Topology: NVLink
Software: TensorRT 10.13, CUDA 13.0

Pinned system specification

llama3.1-8b · Offline: 53,193.6 tokens/s · source

llama3.1-8b · Server: 45,075.8 tokens/s · source