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 24 tok/s/user as the target on DeepSeek V4 Pro 1.6T, GB300 NVL72 produces 11286 tok/s/GPU ($0.07 per million tokens) and MI355X produces 811 ($0.51). GB300 NVL72 is 676% cheaper per token; GB300 NVL72 delivers 1291% more tok/s/GPU.
At 35 tok/s/user interactivity on DeepSeek V4 Pro 1.6T, GB300 NVL72 delivers 9348 tok/s/GPU at $0.08 per million tokens; MI355X delivers 519 tok/s/GPU at $0.79. GB300 NVL72 is 909% cheaper per token; GB300 NVL72 delivers 1700% more tok/s/GPU at this point.
GB300 NVL72 posts 8044 tok/s/GPU for $0.09 per million tokens at 46 tok/s/user on DeepSeek V4 Pro 1.6T; MI355X posts 241 tok/s/GPU for $1.67. GB300 NVL72 is 1735% cheaper per token; GB300 NVL72 delivers 3240% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB300 NVL72:11286.2MI355X:811.5 | GB300 NVL72:9348.0MI355X:519.2 | GB300 NVL72:8044.4MI355X:240.8 |
| Cost ($/M tok) | GB300 NVL72:$0.065MI355X:$0.507 | GB300 NVL72:$0.078MI355X:$0.791 | GB300 NVL72:$0.091MI355X:$1.670 |
| tok/s/MW | GB300 NVL72:5374376MI355X:306224 | GB300 NVL72:4451416MI355X:195925 | GB300 NVL72:3830671MI355X:90885 |
| Concurrency | GB300 NVL72:~12683MI355X:~32 | GB300 NVL72:~2953MI355X:~14 | GB300 NVL72:~1355MI355X:~5 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.