Qwen 3.5 397B-A17B · GPU comparison

Qwen 3.5 397B-A17B — MI355X vs RTX PRO 6000

Head-to-head AI inference benchmark comparison of MI355X (AMD CDNA 4) and RTX PRO 6000 (NVIDIA Blackwell) 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.

Setting 47 tok/s/user as the target on Qwen 3.5 397B-A17B, MI355X produces 5979 tok/s/GPU ($0.07 per million tokens) and RTX PRO 6000 produces 1069 ($0.11). MI355X is 58% cheaper per token; MI355X delivers 459% more tok/s/GPU.

At 82 tok/s/user interactivity on Qwen 3.5 397B-A17B, MI355X delivers 4422 tok/s/GPU at $0.09 per million tokens; RTX PRO 6000 delivers 796 tok/s/GPU at $0.15. MI355X is 57% cheaper per token; MI355X delivers 455% more tok/s/GPU at this point.

MI355X posts 3337 tok/s/GPU for $0.12 per million tokens at 118 tok/s/user on Qwen 3.5 397B-A17B; RTX PRO 6000 posts 538 tok/s/GPU for $0.24. MI355X is 96% cheaper per token; MI355X delivers 520% 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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
MI355X:5979.0RTX PRO 6000:1069.0
MI355X:4422.4RTX PRO 6000:796.2
MI355X:3337.5RTX PRO 6000:537.9
Cost ($/M tok)
MI355X:$0.069RTX PRO 6000:$0.109
MI355X:$0.092RTX PRO 6000:$0.145
MI355X:$0.124RTX PRO 6000:$0.242
tok/s/MW
MI355X:2860763RTX PRO 6000:1096391
MI355X:2115993RTX PRO 6000:816652
MI355X:1596873RTX PRO 6000:551701
Concurrency
MI355X:~29RTX PRO 6000:~12
MI355X:~12RTX PRO 6000:~5
MI355X:~8RTX PRO 6000:~2

Inference Performance

Inference performance metrics across different models, hardware configurations, and serving parameters.

Vendor:
Deployment:
Spec Decoding: