MiniMax M3 428B — B200 vs MI355X
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) 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.
B200 / MI355X on MiniMax M3 428B at 95 tok/s/user: 1518 / 1506 tok/s/GPU, $0.37 / $0.27 per million tokens. MI355X is 37% cheaper per token; throughput per GPU is essentially tied.
Around the middle of the 11–348 tok/s/user interactivity band, at 179 tok/s/user on MiniMax M3 428B: B200 runs 611 tok/s/GPU at $0.89/M tokens, MI355X runs 572 at $0.71/M. MI355X is 25% cheaper per token; B200 delivers 7% more tok/s/GPU.
Setting 264 tok/s/user as the target on MiniMax M3 428B, B200 produces 234 tok/s/GPU ($2.31 per million tokens) and MI355X produces 217 ($1.95). MI355X is 19% cheaper per token; B200 delivers 8% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:1517.9MI355X:1505.7 | B200:610.5MI355X:571.6 | B200:233.7MI355X:217.0 |
| Cost ($/M tok) | B200:$0.373MI355X:$0.273 | B200:$0.887MI355X:$0.708 | B200:$2.313MI355X:$1.945 |
| tok/s/MW | B200:887649MI355X:720453 | B200:357033MI355X:273507 | B200:136664MI355X:103837 |
| Concurrency | B200:~46MI355X:~35 | B200:~7MI355X:~7 | B200:~2MI355X:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.