DeepSeek R1 — H100 vs MI355X
Head-to-head AI inference benchmark comparison of H100 (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.
Near the low end of the 16–128 tok/s/user interactivity band, at 43 tok/s/user on DeepSeek R1: H100 runs 263 tok/s/GPU at $1.38/M tokens, MI355X runs 1573 at $0.26/M. MI355X is 429% cheaper per token; MI355X delivers 498% more tok/s/GPU.
Setting 72 tok/s/user as the target on DeepSeek R1, H100 produces 70 tok/s/GPU ($5.06 per million tokens) and MI355X produces 455 ($0.90). MI355X is 462% cheaper per token; MI355X delivers 551% more tok/s/GPU.
At 100 tok/s/user interactivity on DeepSeek R1, H100 delivers 23 tok/s/GPU at $15.7 per million tokens; MI355X delivers 251 tok/s/GPU at $1.64. MI355X is 857% cheaper per token; MI355X delivers 978% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:262.8MI355X:1572.8 | H100:69.9MI355X:454.8 | H100:23.3MI355X:250.8 |
| Cost ($/M tok) | H100:$1.381MI355X:$0.261 | H100:$5.061MI355X:$0.900 | H100:$15.692MI355X:$1.640 |
| tok/s/MW | H100:151914MI355X:593499 | H100:40376MI355X:171608 | H100:13443MI355X:94630 |
| Concurrency | H100:~529MI355X:~536 | H100:~48MI355X:~43 | H100:~8MI355X:~32 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.