Qwen 3.5 397B-A17B — H100 vs MI300X Performance per Dollar
Cost per million tokens of H100 (NVIDIA Hopper) versus MI300X (AMD CDNA 3) 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.
Push Qwen 3.5 397B-A17B to 52 tok/s/user and H100 lands at $1.20 per million tokens against MI300X's $1.55 — H100 pulls ahead by 29%.
H100: $1.35 per million tokens. MI300X: $2.04. Both at 58 tok/s/user on Qwen 3.5 397B-A17B, with H100 51% cheaper.
Toward the upper edge of the 45–71 tok/s/user interactivity band — at 65 tok/s/user — H100 runs $1.53 per million tokens on Qwen 3.5 397B-A17B while MI300X runs $3.05. H100 is the cheaper choice by 100%. (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 · MI300X $1.12/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:$1.198MI300X:$1.547 | H100:$1.352MI300X:$2.044 | H100:$1.527MI300X:$3.054 |
| Concurrency | H100:~25MI300X:~16 | H100:~19MI300X:~11 | H100:~15MI300X:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.