Qwen 3.5 397B-A17B — H200 vs MI355X Performance per Dollar
Cost per million tokens of H200 (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.
H200: $0.78 per million tokens. MI355X: $0.20. Both at 65 tok/s/user on Qwen 3.5 397B-A17B, with MI355X 297% cheaper.
Around the middle of the 27–182 tok/s/user interactivity band — at 105 tok/s/user — H200 runs $1.10 per million tokens on Qwen 3.5 397B-A17B while MI355X runs $0.34. MI355X is the cheaper choice by 220%.
On Qwen 3.5 397B-A17B at 144 tok/s/user, the per-million math comes out to $1.37 for H200 and $0.52 for MI355X; MI355X delivers 166% more output per dollar. (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): H200 $1.41/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 | H200:$0.775MI355X:$0.195 | H200:$1.096MI355X:$0.342 | H200:$1.374MI355X:$0.516 |
| Concurrency | H200:~31MI355X:~66 | H200:~14MI355X:~23 | H200:~8MI355X:~11 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.