Head-to-head AI inference benchmark comparison of B200 (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.
Near the low end of the 15–53 tok/s/user interactivity band, at 25 tok/s/user on DeepSeek R1: B200 runs 5857 tok/s/chip at $0.08/M tokens, MI325X runs 863 at $0.35/M. B200 is 331% cheaper per token; B200 delivers 578% more tok/s/chip.
Setting 34 tok/s/user as the target on DeepSeek R1, B200 produces 5684 tok/s/chip ($0.08 per million tokens) and MI325X produces 706 ($0.43). B200 is 412% cheaper per token; B200 delivers 705% more tok/s/chip.
At 44 tok/s/user interactivity on DeepSeek R1, B200 delivers 5274 tok/s/chip at $0.09 per million tokens; MI325X delivers 474 tok/s/chip at $0.65. B200 is 608% cheaper per token; B200 delivers 1013% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:5856.9MI325X:863.3 | B200:5683.6MI325X:706.1 | B200:5274.0MI325X:473.7 |
| Cost ($/M tok) | B200:$0.082MI325X:$0.354 | B200:$0.085MI325X:$0.433 | B200:$0.091MI325X:$0.645 |
| tok/s/MW | B200:3425112MI325X:510837 | B200:3323766MI325X:417783 | B200:3084218MI325X:280302 |
| Concurrency | B200:~1088MI325X:~33 | B200:~757MI325X:~20 | B200:~389MI325X:~11 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
DeepSeek R1 0528 671B 8K / 1K Total Tokens per $1 TCO vs. Interactivity