Full guide: run MiniMax H3 fast on AMD GPUs (RX 9070/R9700/7900) with ComfyUI and ROCm 7.14
Apprehensive_Sky892 · reddit · 2026-09-13
A detailed Windows 11 guide for running MiniMax H3 (MMH3) at full speed on AMD GPUs (RDNA 4 RX 9070 / AI Pro R9700, RDNA 3 RX 7900).
The problem: default ComfyUI installs ship PyTorch built on ROCm 7.2, where the int8convrot model fails on these cards; you need ROCm 7.13+ (7.14.0 recommended).
Key command: pip install --index-url https://repo.amd.com/rocm/whl-multi-arch/ "torch[device-gfx????]==2.12.0+rocm7.14.0" — look up your GPU's gfx???? value in AMD's docs.
Four install paths, easiest to hardest: Stability Matrix (author-tested, currently broken — includes full launch args like --enable-dynamic-vram --use-ck-attention), Portable ComfyUI with embedded Python, Python venv + comfy-cli, and manual pip/git setup.
Also includes the Krea-2 int8convrot model link (13.5 GB with SHA256), a verification step, and a promise of an upcoming Ubuntu version of the guide.
More from Infra
- Running MiniMax H3 on AMD GPUs: full ComfyUI + ROCm 7.14 Ubuntu guide — Apprehensive_Sky892 · 2026-09-13
- Why AI infra engineers hate model dynamism: a full-screen scream — generativist · 2026-09-13
- Beff Jezos: big labs are out of compute headroom and engineering panic for a regulatory cartel — beffjezos · 2026-09-13
- Researcher: genuine breakthroughs take orders of magnitude more compute than people realize — generativist · 2026-09-13
- 20-min video backlog cleared in one day on rented 8x RTX PRO 6000 for about $150 — FrontAdhesiveness615 · 2026-09-13
- Yacine quips that Google's Astra is being load-shed — yacineMTB · 2026-09-13