DeepSeek R1 · Chip comparison
DeepSeek R1 — GB300 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB300 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.
Throughput at 95 tok/s/user on DeepSeek R1: GB300 NVL72 hits 12,350 tok/s/chip, MI355X hits 7,592. Costs per million tokens land at $0.052 and $0.055 respectively. GB300 NVL72 is 5% cheaper per token; GB300 NVL72 delivers 63% more tok/s/chip.
GB300 NVL72 / MI355X on DeepSeek R1 at 167 tok/s/user: 4,744 / 1,909 tok/s/chip, $0.14 / $0.22 per million tokens. GB300 NVL72 is 38% cheaper per token; GB300 NVL72 delivers 148% more tok/s/chip.
Toward the upper edge of the 23–312 tok/s/user interactivity band, at 240 tok/s/user on DeepSeek R1: GB300 NVL72 runs 961 tok/s/chip at $0.67 per million tokens, MI355X runs 541 at $0.77. GB300 NVL72 is 13% cheaper per token; GB300 NVL72 delivers 78% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL7212,350MI355X7,592 | GB300 NVL724,744MI355X1,909 | GB300 NVL72961MI355X541 |
| Cost ($/M tok) | GB300 NVL72$0.052MI355X$0.055 | GB300 NVL72$0.14MI355X$0.22 | GB300 NVL72$0.67MI355X$0.77 |
| tok/s/MW | GB300 NVL725,825,707MI355X3,632,663 | GB300 NVL722,237,582MI355X913,385 | GB300 NVL72453,449MI355X258,638 |
| Concurrency | GB300 NVL72~690MI355X~204 | GB300 NVL72~181MI355X~55 | GB300 NVL72~23MI355X~8 |