Qwen 3.5 397B-A17B — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on Qwen 3.5 397B-A17B. 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.
At 64 tok/s/user interactivity on Qwen 3.5 397B-A17B, GB200 NVL72 delivers 1239 tok/s/GPU at $0.51 per million tokens; MI355X delivers 2131 tok/s/GPU at $0.19. MI355X is 163% cheaper per token; MI355X delivers 72% more tok/s/GPU at this point.
GB200 NVL72 posts 464 tok/s/GPU for $1.30 per million tokens at 101 tok/s/user on Qwen 3.5 397B-A17B; MI355X posts 1218 tok/s/GPU for $0.33. MI355X is 293% cheaper per token; MI355X delivers 163% more tok/s/GPU.
Throughput at 137 tok/s/user on Qwen 3.5 397B-A17B: GB200 NVL72 hits 206 tok/s/GPU, MI355X hits 873. Per-million costs land at $2.98 and $0.47 respectively. MI355X is 529% cheaper per token; MI355X delivers 324% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | GB200 NVL72:1239.3MI355X:2131.3 | GB200 NVL72:463.7MI355X:1217.8 | GB200 NVL72:205.8MI355X:872.6 |
| Cost ($/M tok) | GB200 NVL72:$0.506MI355X:$0.193 | GB200 NVL72:$1.297MI355X:$0.330 | GB200 NVL72:$2.983MI355X:$0.475 |
| tok/s/MW | GB200 NVL72:590129MI355X:804277 | GB200 NVL72:220809MI355X:459536 | GB200 NVL72:98013MI355X:329279 |
| Concurrency | GB200 NVL72:~113MI355X:~68 | GB200 NVL72:~20MI355X:~25 | GB200 NVL72:~7MI355X:~13 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.