DeepSeek R1 — B200 vs MI355X
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) 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.
B200 posts 2916 tok/s/GPU for $0.19 per million tokens at 58 tok/s/user on DeepSeek R1; MI355X posts 1603 tok/s/GPU for $0.26. B200 is 38% cheaper per token; B200 delivers 82% more tok/s/GPU.
Throughput at 107 tok/s/user on DeepSeek R1: B200 hits 1298 tok/s/GPU, MI355X hits 679. Per-million costs land at $0.42 and $0.61 respectively. B200 is 46% cheaper per token; B200 delivers 91% more tok/s/GPU.
B200 / MI355X on DeepSeek R1 at 156 tok/s/user: 756 / 419 tok/s/GPU, $0.72 / $0.98 per million tokens. B200 is 37% cheaper per token; B200 delivers 80% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:2916.3MI355X:1603.2 | B200:1298.1MI355X:678.5 | B200:755.9MI355X:419.1 |
| Cost ($/M tok) | B200:$0.186MI355X:$0.256 | B200:$0.418MI355X:$0.608 | B200:$0.717MI355X:$0.981 |
| tok/s/MW | B200:1343910MI355X:604998 | B200:598183MI355X:256052 | B200:348348MI355X:158169 |
| Concurrency | B200:~284MI355X:~37 | B200:~78MI355X:~11 | B200:~60MI355X:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.