MI355X FP4: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on MI355X FP4 (AMD CDNA 4) running Qwen 3.5 397B-A17B. Throughput, cost, and interactivity differences across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
MTP acceptance-rate implementations differ across inference engines. Points from different engines are not directly comparable on the same curve — throughput and cost at matched interactivity may reflect engine-level differences rather than pure speculative decoding gains. Interpret cross-engine comparisons with caution.
MTP posts 2351 tok/s/GPU for $0.17 per million tokens at 53 tok/s/user on Qwen 3.5 397B-A17B (MI355X FP4); Off posts 1505 tok/s/GPU for $0.27. MTP is 56% cheaper per token; MTP delivers 56% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 77 tok/s/user on Qwen 3.5 397B-A17B (MI355X FP4): MTP hits 1963 tok/s/GPU, Off hits 909. Per-million costs land at $0.21 and $0.44 respectively. MTP is 107% cheaper per token; MTP delivers 116% more tok/s/GPU. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size.
Toward the upper edge of the 30–123 tok/s/user interactivity band, at 100 tok/s/user on Qwen 3.5 397B-A17B (MI355X FP4): MTP runs 1459 tok/s/GPU at $0.28/M tokens, Off runs 563 at $0.73/M. MTP is 159% cheaper per token; MTP delivers 159% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less. (Numbers reflect this URL's pinned 1k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:2350.7Off:1505.1 | MTP:1963.5Off:909.0 | MTP:1458.6Off:562.9 |
| Cost ($/M tok) | MTP:$0.175Off:$0.273 | MTP:$0.212Off:$0.439 | MTP:$0.281Off:$0.728 |
| tok/s/MW | MTP:1124716Off:720132 | MTP:939462Off:434913 | MTP:697900Off:269311 |
| Concurrency | MTP:~45Off:~29 | MTP:~26Off:~11 | MTP:~15Off:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.