DeepSeek R1 — GB200 NVL72 vs MI325X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) 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 37 tok/s/user on DeepSeek R1: GB200 NVL72 hits 7073 tok/s/chip, MI325X hits 646. Per-million costs land at $0.07 and $0.47 respectively. GB200 NVL72 is 543% cheaper per token; GB200 NVL72 delivers 994% more tok/s/chip.
GB200 NVL72 / MI325X on DeepSeek R1 at 43 tok/s/user: 6988 / 500 tok/s/chip, $0.07 / $0.60 per million tokens. GB200 NVL72 is 709% cheaper per token; GB200 NVL72 delivers 1297% more tok/s/chip.
Toward the upper edge of the 32–53 tok/s/user interactivity band, at 48 tok/s/user on DeepSeek R1: GB200 NVL72 runs 6861 tok/s/chip at $0.08/M tokens, MI325X runs 371 at $0.83/M. GB200 NVL72 is 1005% cheaper per token; GB200 NVL72 delivers 1750% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:7072.6MI325X:646.3 | GB200 NVL72:6988.1MI325X:500.1 | GB200 NVL72:6861.5MI325X:370.9 |
| Cost ($/M tok) | GB200 NVL72:$0.073MI325X:$0.470 | GB200 NVL72:$0.074MI325X:$0.598 | GB200 NVL72:$0.075MI325X:$0.832 |
| tok/s/MW | GB200 NVL72:3782124MI325X:382442 | GB200 NVL72:3736945MI325X:295910 | GB200 NVL72:3669226MI325X:219492 |
| Concurrency | GB200 NVL72:~2484MI325X:~17 | GB200 NVL72:~4883MI325X:~11 | GB200 NVL72:~6142MI325X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.