Head-to-head AI inference benchmark comparison of MI325X (AMD CDNA 3) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on DeepSeekv4 Pro 0813 1.6T. 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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Vera Rubin NVL72: 170118 tok/s/chip, $0.01 per million tokens at 125 tok/s/user on DeepSeekv4 Pro 0813 1.6T. MI325X is unmeasured here.
At 175 tok/s/user on DeepSeekv4 Pro 0813 1.6T, Vera Rubin NVL72 delivers 61937 tok/s/chip at $0.02 per million tokens; MI325X hasn't been benchmarked at this target.
Vera Rubin NVL72 hits 7223 tok/s/chip for $0.14 per million tokens at 226 tok/s/user on DeepSeekv4 Pro 0813 1.6T. No MI325X data at this operating point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | MI325X:—Vera Rubin NVL72:170117.7 | MI325X:—Vera Rubin NVL72:61936.7 | MI325X:—Vera Rubin NVL72:7223.3 |
| Cost ($/M tok) | MI325X:—Vera Rubin NVL72:$0.006 | MI325X:—Vera Rubin NVL72:$0.016 | MI325X:—Vera Rubin NVL72:$0.139 |
| tok/s/MW | MI325X:—Vera Rubin NVL72:51550815 | MI325X:—Vera Rubin NVL72:18768687 | MI325X:—Vera Rubin NVL72:2188894 |
| Concurrency | MI325X:—Vera Rubin NVL72:~1178 | MI325X:—Vera Rubin NVL72:~547 | MI325X:—Vera Rubin NVL72:~33 |
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.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
DeepSeek V4 Pro 0813 1.6T 8K / 1K Total Tokens per $1 TCO vs. Interactivity