Qwen 3.5 397B-A17B — H100 vs MI355X Performance per Dollar
Cost per million tokens of H100 (NVIDIA Hopper) versus MI355X (AMD CDNA 4) on Qwen 3.5 397B-A17B. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Near the low end of the 29–165 tok/s/user interactivity band — at 63 tok/s/user — H100 runs $0.73 per million tokens on Qwen 3.5 397B-A17B while MI355X runs $0.19. MI355X is the cheaper choice by 283%.
On Qwen 3.5 397B-A17B at 97 tok/s/user, the per-million math comes out to $1.66 for H100 and $0.32 for MI355X; MI355X delivers 424% more output per dollar.
At 131 tok/s/user on Qwen 3.5 397B-A17B, H100 costs $4.06 per million tokens; MI355X costs $0.44. MI355X is 818% more cost-efficient at this operating point. (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.)
GPU pricing (owning hyperscaler): H100 $1.30/GPU/hr · MI355X $1.48/GPU/hr. Source: SemiAnalysis Market August 2025 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | H100:$0.726MI355X:$0.190 | H100:$1.664MI355X:$0.318 | H100:$4.057MI355X:$0.442 |
| Concurrency | H100:~39MI355X:~69 | H100:~10MI355X:~27 | H100:~3MI355X:~14 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.