DeepSeek R1 — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (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.
Setting 74 tok/s/user as the target on DeepSeek R1, GB200 NVL72 produces 7804 tok/s/GPU ($0.08 per million tokens) and MI355X produces 1767 ($0.23). GB200 NVL72 is 196% cheaper per token; GB200 NVL72 delivers 342% more tok/s/GPU.
At 131 tok/s/user interactivity on DeepSeek R1, GB200 NVL72 delivers 1693 tok/s/GPU at $0.38 per million tokens; MI355X delivers 418 tok/s/GPU at $1.00. GB200 NVL72 is 161% cheaper per token; GB200 NVL72 delivers 305% more tok/s/GPU at this point.
GB200 NVL72 posts 655 tok/s/GPU for $1.05 per million tokens at 188 tok/s/user on DeepSeek R1; MI355X posts 181 tok/s/GPU for $2.28. GB200 NVL72 is 118% cheaper per token; GB200 NVL72 delivers 263% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:7803.7MI355X:1767.2 | GB200 NVL72:1692.9MI355X:417.7 | GB200 NVL72:655.0MI355X:180.7 |
| Cost ($/M tok) | GB200 NVL72:$0.079MI355X:$0.233 | GB200 NVL72:$0.383MI355X:$0.997 | GB200 NVL72:$1.047MI355X:$2.283 |
| tok/s/MW | GB200 NVL72:3716055MI355X:666886 | GB200 NVL72:806147MI355X:157623 | GB200 NVL72:311890MI355X:68172 |
| Concurrency | GB200 NVL72:~2211MI355X:~254 | GB200 NVL72:~311MI355X:~22 | GB200 NVL72:~96MI355X:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.