llamaperf

H100 80GB vs AMD Strix Halo 128GB

For running local LLMs · 6 reports across 3 models

Side A
H100 80GB
Vendor
nvidia
VRAM
80GB
Memory
Discrete
Side B
AMD Strix Halo 128GB
Vendor
amd
VRAM
128GB
Memory
Unified

Tokens per second by model

ModelH100 80GBAMD Strix Halo 128GB
Qwen3.6up to 35B45.0n=121.2n=3
Gemma 4up to 31B125.3n=1
Qwen3.5up to 27Bn=1

More comparisons