DeepSeek R1 — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (AMD CDNA 3) 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.
H200 posts 1993 tok/s/GPU for $0.19 per million tokens at 36 tok/s/user on DeepSeek R1; MI300X posts 268 tok/s/GPU for $1.16. H200 is 501% cheaper per token; H200 delivers 644% more tok/s/GPU.
Throughput at 47 tok/s/user on DeepSeek R1: H200 hits 915 tok/s/GPU, MI300X hits 186. Per-million costs land at $0.42 and $1.66 respectively. H200 is 293% cheaper per token; H200 delivers 391% more tok/s/GPU.
H200 / MI300X on DeepSeek R1 at 58 tok/s/user: 427 / 113 tok/s/GPU, $0.92 / $2.74 per million tokens. H200 is 198% cheaper per token; H200 delivers 277% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1992.5MI300X:267.6 | H200:915.2MI300X:186.4 | H200:426.8MI300X:113.3 |
| Cost ($/M tok) | H200:$0.194MI300X:$1.163 | H200:$0.423MI300X:$1.663 | H200:$0.919MI300X:$2.737 |
| tok/s/MW | H200:1151755MI300X:149525 | H200:529007MI300X:104145 | H200:246710MI300X:63313 |
| Concurrency | H200:~1024MI300X:~31 | H200:~862MI300X:~17 | H200:~66MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.