Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on MiniMax M2.5/M2.7. 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 48 tok/s/user as the target on MiniMax M2.5/M2.7, GB200 NVL72 produces 8970 tok/s/chip ($0.06 per million tokens) and MI325X produces 2106 ($0.15). GB200 NVL72 is 152% cheaper per token; GB200 NVL72 delivers 326% more tok/s/chip.
At 65 tok/s/user interactivity on MiniMax M2.5/M2.7, GB200 NVL72 delivers 5557 tok/s/chip at $0.09 per million tokens; MI325X delivers 1470 tok/s/chip at $0.21. GB200 NVL72 is 124% cheaper per token; GB200 NVL72 delivers 278% more tok/s/chip at this point.
GB200 NVL72 posts 3194 tok/s/chip for $0.16 per million tokens at 82 tok/s/user on MiniMax M2.5/M2.7; MI325X posts 878 tok/s/chip for $0.35. GB200 NVL72 is 115% cheaper per token; GB200 NVL72 delivers 264% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:8969.9MI325X:2106.1 | GB200 NVL72:5556.6MI325X:1470.3 | GB200 NVL72:3194.4MI325X:878.4 |
| Cost ($/M tok) | GB200 NVL72:$0.058MI325X:$0.145 | GB200 NVL72:$0.093MI325X:$0.208 | GB200 NVL72:$0.162MI325X:$0.348 |
| tok/s/MW | GB200 NVL72:4796740MI325X:1246204 | GB200 NVL72:2971457MI325X:869998 | GB200 NVL72:1708227MI325X:519748 |
| Concurrency | GB200 NVL72:~146MI325X:~10 | GB200 NVL72:~64MI325X:~5 | GB200 NVL72:~28MI325X:~7 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr: VR200: 3.61H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68Jalapeño (Teacup): 1.47
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
