llamaperf

H100 80GB vs RX 9070

For running local LLMs · 3 reports across 2 models

Side A
H100 80GB
Vendor
nvidia
VRAM
80GB
Memory
Discrete
Side B
RX 9070
Vendor
amd
VRAM
16GB
Memory
Discrete

Tokens per second by model

ModelH100 80GBRX 9070
Qwen3.6up to 35B45.0n=146.9n=1
Gemma 4up to 31B125.3n=1

More comparisons