DeepSeek V4 Pro 1.6T — GB300 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on DeepSeek V4 Pro 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.
Setting 47 tok/s/user as the target on DeepSeek V4 Pro 1.6T, GB300 NVL72 produces 11474 tok/s/GPU ($0.06 per million tokens) and MI355X produces 1413 ($0.29). GB300 NVL72 is 353% cheaper per token; GB300 NVL72 delivers 712% more tok/s/GPU.
At 82 tok/s/user interactivity on DeepSeek V4 Pro 1.6T, GB300 NVL72 delivers 7963 tok/s/GPU at $0.09 per million tokens; MI355X delivers 702 tok/s/GPU at $0.59. GB300 NVL72 is 533% cheaper per token; GB300 NVL72 delivers 1034% more tok/s/GPU at this point.
GB300 NVL72 posts 4460 tok/s/GPU for $0.17 per million tokens at 116 tok/s/user on DeepSeek V4 Pro 1.6T; MI355X posts 262 tok/s/GPU for $1.60. GB300 NVL72 is 867% cheaper per token; GB300 NVL72 delivers 1600% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB300 NVL72:11474.0MI355X:1412.7 | GB300 NVL72:7962.6MI355X:702.2 | GB300 NVL72:4459.5MI355X:262.3 |
| Cost ($/M tok) | GB300 NVL72:$0.064MI355X:$0.291 | GB300 NVL72:$0.092MI355X:$0.585 | GB300 NVL72:$0.166MI355X:$1.600 |
| tok/s/MW | GB300 NVL72:5463799MI355X:533104 | GB300 NVL72:3791714MI355X:264979 | GB300 NVL72:2123574MI355X:98962 |
| Concurrency | GB300 NVL72:~3561MI355X:~27 | GB300 NVL72:~606MI355X:~8 | GB300 NVL72:~256MI355X:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.