Qwen 3.5 397B-A17B — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) on Qwen 3.5 397B-A17B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
At 65 tok/s/user interactivity on Qwen 3.5 397B-A17B, H200 delivers 505 tok/s/GPU at $0.78 per million tokens; MI355X delivers 2102 tok/s/GPU at $0.20. MI355X is 297% cheaper per token; MI355X delivers 316% more tok/s/GPU at this point.
H200 posts 353 tok/s/GPU for $1.10 per million tokens at 105 tok/s/user on Qwen 3.5 397B-A17B; MI355X posts 1172 tok/s/GPU for $0.34. MI355X is 220% cheaper per token; MI355X delivers 232% more tok/s/GPU.
Throughput at 144 tok/s/user on Qwen 3.5 397B-A17B: H200 hits 284 tok/s/GPU, MI355X hits 804. Per-million costs land at $1.37 and $0.52 respectively. MI355X is 166% cheaper per token; MI355X delivers 183% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:505.3MI355X:2101.9 | H200:353.3MI355X:1172.1 | H200:284.4MI355X:804.0 |
| Cost ($/M tok) | H200:$0.775MI355X:$0.195 | H200:$1.096MI355X:$0.342 | H200:$1.374MI355X:$0.516 |
| tok/s/MW | H200:292068MI355X:793161 | H200:204204MI355X:442289 | H200:164386MI355X:303401 |
| Concurrency | H200:~31MI355X:~66 | H200:~14MI355X:~23 | H200:~8MI355X:~11 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.