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 5857 tok/s/chip ($0.08 per million tokens) and MI300X produces 725 ($0.36). B200 is 343% cheaper per token; B200 delivers 708% more tok/s/chip.
At 32 tok/s/user interactivity on DeepSeek R1, B200 delivers 5738 tok/s/chip at $0.08 per million tokens; MI300X delivers 601 tok/s/chip at $0.44. B200 is 424% cheaper per token; B200 delivers 855% more tok/s/chip at this point.
B200 posts 5380 tok/s/chip for $0.09 per million tokens at 42 tok/s/user on DeepSeek R1; MI300X posts 406 tok/s/chip for $0.65. B200 is 627% cheaper per token; B200 delivers 1224% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:5856.9MI300X:725.2 | B200:5738.5MI300X:601.1 | B200:5380.3MI300X:406.3 |
| Cost ($/M tok) | B200:$0.082MI300X:$0.364 | B200:$0.084MI300X:$0.439 | B200:$0.089MI300X:$0.650 |
| tok/s/MW | B200:3425112MI300X:521723 | B200:3355839MI300X:432443 | B200:3146368MI300X:292288 |
| Concurrency | B200:~1088MI300X:~30 | B200:~850MI300X:~18 | B200:~445MI300X:~10 |
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