DeepSeek R1 — H200 vs MI325X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (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.
Throughput at 39 tok/s/user on DeepSeek R1: H200 hits 1747 tok/s/GPU, MI325X hits 341. Per-million costs land at $0.22 and $1.03 respectively. H200 is 362% cheaper per token; H200 delivers 412% more tok/s/GPU.
H200 / MI325X on DeepSeek R1 at 50 tok/s/user: 818 / 225 tok/s/GPU, $0.47 / $1.57 per million tokens. H200 is 230% cheaper per token; H200 delivers 263% more tok/s/GPU.
Toward the upper edge of the 29–70 tok/s/user interactivity band, at 60 tok/s/user on DeepSeek R1: H200 runs 419 tok/s/GPU at $0.94/M tokens, MI325X runs 132 at $2.67/M. H200 is 185% cheaper per token; H200 delivers 217% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1747.1MI325X:341.4 | H200:818.5MI325X:225.5 | H200:419.1MI325X:132.2 |
| Cost ($/M tok) | H200:$0.223MI325X:$1.027 | H200:$0.475MI325X:$1.566 | H200:$0.936MI325X:$2.672 |
| tok/s/MW | H200:1009908MI325X:156612 | H200:473121MI325X:103433 | H200:242262MI325X:60648 |
| Concurrency | H200:~1024MI325X:~39 | H200:~772MI325X:~19 | H200:~215MI325X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.