DeepSeek R1 — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) on DeepSeek R1. 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.
Setting 59 tok/s/user as the target on DeepSeek R1, H200 produces 421 tok/s/chip ($0.80 per million tokens) and MI355X produces 920 ($0.45). MI355X is 79% cheaper per token; MI355X delivers 119% more tok/s/chip.
At 98 tok/s/user interactivity on DeepSeek R1, H200 delivers 152 tok/s/chip at $2.30 per million tokens; MI355X delivers 455 tok/s/chip at $0.92. MI355X is 151% cheaper per token; MI355X delivers 199% more tok/s/chip at this point.
H200 posts 40 tok/s/chip for $8.60 per million tokens at 136 tok/s/user on DeepSeek R1; MI355X posts 217 tok/s/chip for $1.91. MI355X is 351% cheaper per token; MI355X delivers 449% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:420.5MI355X:920.4 | H200:152.1MI355X:454.5 | H200:39.6MI355X:217.1 |
| Cost ($/M tok) | H200:$0.805MI355X:$0.450 | H200:$2.304MI355X:$0.917 | H200:$8.600MI355X:$1.906 |
| tok/s/MW | H200:306952MI355X:440389 | H200:111034MI355X:217476 | H200:28878MI355X:103898 |
| Concurrency | H200:~133MI355X:~199 | H200:~6MI355X:~18 | H200:~14MI355X:~6 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.