MiniMax M3 428B — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) on MiniMax M3 428B. 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.
H200 posts 1454 tok/s/GPU for $0.27 per million tokens at 18 tok/s/user on MiniMax M3 428B; MI355X posts 2324 tok/s/GPU for $0.18. MI355X is 52% cheaper per token; MI355X delivers 60% more tok/s/GPU.
Throughput at 27 tok/s/user on MiniMax M3 428B: H200 hits 1408 tok/s/GPU, MI355X hits 1934. Per-million costs land at $0.28 and $0.21 respectively. MI355X is 33% cheaper per token; MI355X delivers 37% more tok/s/GPU.
H200 / MI355X on MiniMax M3 428B at 37 tok/s/user: 1318 / 1514 tok/s/GPU, $0.30 / $0.28 per million tokens. MI355X is 5% cheaper per token; MI355X delivers 15% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1454.3MI355X:2324.1 | H200:1408.3MI355X:1933.9 | H200:1318.3MI355X:1513.8 |
| Cost ($/M tok) | H200:$0.269MI355X:$0.177 | H200:$0.278MI355X:$0.209 | H200:$0.297MI355X:$0.281 |
| tok/s/MW | H200:840643MI355X:877033 | H200:814056MI355X:729789 | H200:762039MI355X:571247 |
| Concurrency | H200:~184MI355X:~294 | H200:~111MI355X:~158 | H200:~77MI355X:~97 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.