MiniMax M3 428B — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (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.
Setting 66 tok/s/user as the target on MiniMax M3 428B, GB200 NVL72 produces 4050 tok/s/GPU ($0.15 per million tokens) and MI355X produces 4945 ($0.08). MI355X is 82% cheaper per token; MI355X delivers 22% more tok/s/GPU.
At 106 tok/s/user interactivity on MiniMax M3 428B, GB200 NVL72 delivers 1912 tok/s/GPU at $0.32 per million tokens; MI355X delivers 3573 tok/s/GPU at $0.11. MI355X is 184% cheaper per token; MI355X delivers 87% more tok/s/GPU at this point.
GB200 NVL72 posts 811 tok/s/GPU for $0.78 per million tokens at 147 tok/s/user on MiniMax M3 428B; MI355X posts 2481 tok/s/GPU for $0.17. MI355X is 367% cheaper per token; MI355X delivers 206% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:4050.0MI355X:4945.2 | GB200 NVL72:1912.4MI355X:3573.5 | GB200 NVL72:811.5MI355X:2480.7 |
| Cost ($/M tok) | GB200 NVL72:$0.152MI355X:$0.083 | GB200 NVL72:$0.325MI355X:$0.115 | GB200 NVL72:$0.777MI355X:$0.166 |
| tok/s/MW | GB200 NVL72:2165792MI355X:2366116 | GB200 NVL72:1022690MI355X:1709792 | GB200 NVL72:433944MI355X:1186915 |
| Concurrency | GB200 NVL72:~222MI355X:~40 | GB200 NVL72:~45MI355X:~18 | GB200 NVL72:~13MI355X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.