DeepSeek V4 Pro 1.6T — B300 vs MI355X
Head-to-head AI inference benchmark comparison of B300 (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.
B300 posts 5182 tok/s/GPU for $0.13 per million tokens at 40 tok/s/user on DeepSeek V4 Pro 1.6T; MI355X posts 1673 tok/s/GPU for $0.25. B300 is 97% cheaper per token; B300 delivers 210% more tok/s/GPU.
Throughput at 77 tok/s/user on DeepSeek V4 Pro 1.6T: B300 hits 2157 tok/s/GPU, MI355X hits 783. Per-million costs land at $0.30 and $0.52 respectively. B300 is 74% cheaper per token; B300 delivers 175% more tok/s/GPU.
B300 / MI355X on DeepSeek V4 Pro 1.6T at 114 tok/s/user: 1464 / 309 tok/s/GPU, $0.44 / $1.42 per million tokens. B300 is 219% cheaper per token; B300 delivers 373% more tok/s/GPU. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:5182.1MI355X:1672.6 | B300:2156.8MI355X:783.5 | B300:1463.6MI355X:309.3 |
| Cost ($/M tok) | B300:$0.126MI355X:$0.247 | B300:$0.301MI355X:$0.525 | B300:$0.444MI355X:$1.416 |
| tok/s/MW | B300:2388072MI355X:631180 | B300:993908MI355X:295658 | B300:674488MI355X:116708 |
| Concurrency | B300:~1024MI355X:~68 | B300:~15MI355X:~10 | B300:~7MI355X:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.