MI355X FP4: MTP vs Off Speculative Decoding
Speculative decoding comparison of MTP versus Off on MI355X FP4 (AMD CDNA 4) 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 4–75 tok/s/user interactivity band, at 21 tok/s/user on DeepSeek V4 Pro 1.6T (MI355X FP4): MTP runs 5202 tok/s/chip at $0.08/M tokens, Off runs 4487 at $0.09/M. MTP is 16% cheaper per token; MTP delivers 16% more tok/s/chip. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 39 tok/s/user on DeepSeek V4 Pro 1.6T (MI355X FP4), MTP delivers 3789 tok/s/chip at $0.11 per million tokens; Off delivers 1330 tok/s/chip at $0.31. MTP is 185% cheaper per token; MTP delivers 185% more tok/s/chip. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution.
MTP posts 2198 tok/s/chip for $0.19 per million tokens at 57 tok/s/user on DeepSeek V4 Pro 1.6T (MI355X FP4); Off posts 593 tok/s/chip for $0.70. MTP is 270% cheaper per token; MTP delivers 270% more tok/s/chip. Draft-token acceptance rates determine whether speculative decoding helps or hurts at a given concurrency level. (Numbers reflect this URL's pinned 8k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | MTP:5201.9Off:4487.0 | MTP:3789.4Off:1329.6 | MTP:2197.8Off:593.3 |
| Cost ($/M tok) | MTP:$0.080Off:$0.093 | MTP:$0.110Off:$0.313 | MTP:$0.190Off:$0.702 |
| tok/s/MW | MTP:2488936Off:2146870 | MTP:1813116Off:636183 | MTP:1051575Off:283856 |
| Concurrency | MTP:~257Off:~916 | MTP:~291Off:~16 | MTP:~47Off:~10 |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.