MiniMax M3 428B — MI300X vs MI355X
Head-to-head AI inference benchmark comparison of MI300X (AMD CDNA 3) 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.
At 45 tok/s/user interactivity on MiniMax M3 428B, MI300X delivers 707 tok/s/chip at $0.37 per million tokens; MI355X delivers 6223 tok/s/chip at $0.07. MI355X is 456% cheaper per token; MI355X delivers 780% more tok/s/chip at this point.
MI300X posts 588 tok/s/chip for $0.45 per million tokens at 84 tok/s/user on MiniMax M3 428B; MI355X posts 4187 tok/s/chip for $0.10. MI355X is 354% cheaper per token; MI355X delivers 612% more tok/s/chip.
Throughput at 123 tok/s/user on MiniMax M3 428B: MI300X hits 352 tok/s/chip, MI355X hits 3108. Per-million costs land at $0.73 and $0.13 respectively. MI355X is 444% cheaper per token; MI355X delivers 783% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | MI300X:707.5MI355X:6222.6 | MI300X:587.9MI355X:4186.6 | MI300X:352.0MI355X:3108.4 |
| Cost ($/M tok) | MI300X:$0.373MI355X:$0.067 | MI300X:$0.449MI355X:$0.099 | MI300X:$0.733MI355X:$0.135 |
| tok/s/MW | MI300X:508971MI355X:2977297 | MI300X:422953MI355X:2003165 | MI300X:253227MI355X:1487295 |
| Concurrency | MI300X:~16MI355X:~97 | MI300X:~7MI355X:~26 | MI300X:~3MI355X:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.