DeepSeek R1 — B200 vs MI355X Performance per Dollar
Cost per million tokens of B200 (NVIDIA Blackwell) versus MI355X (AMD CDNA 4) on DeepSeek R1. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Push DeepSeek R1 to 87 tok/s/user and B200 lands at $0.06 per million tokens against MI355X's $0.05 — MI355X pulls ahead by 14%.
B200: $0.23 per million tokens. MI355X: $0.19. Both at 158 tok/s/user on DeepSeek R1, with MI355X 21% cheaper.
Toward the upper edge of the 16–300 tok/s/user interactivity band — at 230 tok/s/user — B200 runs $0.38 per million tokens on DeepSeek R1 while MI355X runs $0.72. B200 is the cheaper choice by 87%. (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.)
Chip pricing (owning hyperscaler): B200 $1.73/chip/hr · MI355X $1.50/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | B200:$0.060MI355X:$0.053 | B200:$0.234MI355X:$0.194 | B200:$0.383MI355X:$0.717 |
| Concurrency | B200:~497MI355X:~254 | B200:~62MI355X:~64 | B200:~30MI355X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.