MI325X FP8: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on MI325X FP8 (AMD CDNA 3) running DeepSeek R1. 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 334 tok/s/GPU for $1.06 per million tokens at 39 tok/s/user on DeepSeek R1 (MI325X FP8); Off posts 346 tok/s/GPU for $1.01. Off is 4% cheaper per token; Off delivers 4% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level.
Throughput at 50 tok/s/user on DeepSeek R1 (MI325X FP8): MTP hits 219 tok/s/GPU, Off hits 226. Per-million costs land at $1.61 and $1.56 respectively. Off is 3% cheaper per token; Off delivers 3% 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 29–70 tok/s/user interactivity band, at 60 tok/s/user on DeepSeek R1 (MI325X FP8): MTP runs 132 tok/s/GPU at $2.68/M tokens, Off runs 131 at $2.70/M. MTP is 1% cheaper per token; throughput per GPU is essentially tied. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less. (Numbers reflect this URL's pinned 1k/1k · fp8 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:334.0Off:346.0 | MTP:218.9Off:225.6 | MTP:131.6Off:130.8 |
| Cost ($/M tok) | MTP:$1.056Off:$1.012 | MTP:$1.611Off:$1.563 | MTP:$2.676Off:$2.704 |
| tok/s/MW | MTP:153222Off:158705 | MTP:100422Off:103480 | MTP:60383Off:60012 |
| Concurrency | MTP:~36Off:~38 | MTP:~18Off:~19 | MTP:~9Off:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.