MI300X FP8: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on MI300X FP8 (AMD CDNA 3) running DeepSeek V4 Pro 1.6T. 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.
Near the low end of the 5–20 tok/s/user interactivity band, at 9 tok/s/user on DeepSeek V4 Pro 1.6T (MI300X FP8): MTP runs 370 tok/s/GPU at $0.83/M tokens, Off runs 243 at $1.27/M. MTP is 54% cheaper per token; MTP delivers 52% more tok/s/GPU. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 13 tok/s/user on DeepSeek V4 Pro 1.6T (MI300X FP8), MTP delivers 263 tok/s/GPU at $1.19 per million tokens; Off delivers 135 tok/s/GPU at $2.29. MTP is 93% cheaper per token; MTP delivers 95% more tok/s/GPU. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution.
MTP posts 163 tok/s/GPU for $1.94 per million tokens at 17 tok/s/user on DeepSeek V4 Pro 1.6T (MI300X FP8); Off posts 67 tok/s/GPU for $4.63. MTP is 139% cheaper per token; MTP delivers 143% more tok/s/GPU. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level. (Numbers reflect this URL's pinned 1k/1k · fp8 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | MTP:370.4Off:243.1 | MTP:262.5Off:134.6 | MTP:162.9Off:67.0 |
| Cost ($/M tok) | MTP:$0.825Off:$1.274 | MTP:$1.185Off:$2.291 | MTP:$1.941Off:$4.633 |
| tok/s/MW | MTP:266449Off:174897 | MTP:188861Off:96819 | MTP:117202Off:48233 |
| Concurrency | MTP:~170Off:~109 | MTP:~84Off:~42 | MTP:~39Off:~16 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.