Qwen 3.5 397B-A17B — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (AMD CDNA 3) 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.
Throughput at 43 tok/s/user on Qwen 3.5 397B-A17B: H200 hits 626 tok/s/GPU, MI300X hits 346. Per-million costs land at $0.62 and $0.89 respectively. H200 is 44% cheaper per token; H200 delivers 81% more tok/s/GPU.
H200 / MI300X on Qwen 3.5 397B-A17B at 53 tok/s/user: 558 / 190 tok/s/GPU, $0.69 / $1.63 per million tokens. H200 is 135% cheaper per token; H200 delivers 195% more tok/s/GPU.
Toward the upper edge of the 34–71 tok/s/user interactivity band, at 62 tok/s/user on Qwen 3.5 397B-A17B: H200 runs 520 tok/s/GPU at $0.75/M tokens, MI300X runs 123 at $2.52/M. H200 is 234% cheaper per token; H200 delivers 324% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:626.3MI300X:345.8 | H200:558.3MI300X:189.5 | H200:520.1MI300X:122.7 |
| Cost ($/M tok) | H200:$0.622MI300X:$0.893 | H200:$0.694MI300X:$1.633 | H200:$0.753MI300X:$2.517 |
| tok/s/MW | H200:362024MI300X:193162 | H200:322741MI300X:105886 | H200:300618MI300X:68573 |
| Concurrency | H200:~58MI300X:~34 | H200:~42MI300X:~15 | H200:~34MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.