DeepSeek R1 · Chip comparison
DeepSeek R1 — B200 vs MI300X
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and MI300X (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.
Setting 23 tok/s/user as the target on DeepSeek R1, B200 produces 5,857 tok/s/chip ($0.082 per million tokens) and MI300X produces 706 ($0.37). B200 is 78% cheaper per token; B200 delivers 8.3× the tok/s/chip.
At 33 tok/s/user interactivity on DeepSeek R1, B200 delivers 5,712 tok/s/chip at $0.084 per million tokens; MI300X delivers 592 tok/s/chip at $0.45. B200 is 81% cheaper per token; B200 delivers 9.6× the tok/s/chip at this point.
B200 posts 5,274 tok/s/chip for $0.091 per million tokens at 44 tok/s/user on DeepSeek R1; MI300X posts 407 tok/s/chip for $0.65. B200 is 86% cheaper per token; B200 delivers 13× the tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B2005,857MI300X706 | B2005,712MI300X592 | B2005,274MI300X407 |
| Cost ($/M tok) | B200$0.082MI300X$0.37 | B200$0.084MI300X$0.45 | B200$0.091MI300X$0.65 |
| tok/s/MW | B2003,425,112MI300X508,108 | B2003,340,209MI300X426,187 | B2003,084,218MI300X293,040 |
| Concurrency | B200~1,088MI300X~28 | B200~803MI300X~17 | B200~389MI300X~9 |