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.
Related event: Guide: running MiniMax H3 on AMD GPUs with ComfyUI and ROCm(2 posts)→
More from Infra
- Clover-35B-A3B Quantized to 29.4 GB Runs at ~42 tok/s on a 64 GB Mac — viveky259 · 2026-10-31
- Liquid Agent brings fully local LFM3 agent to Linux with one-command install — JosephJacks_ · 2026-10-31
- A fully local office AI stack: MiniCPM, Qwen3, LFM2.5 and Whisper — HowDevelop · 2026-10-31
- a16z: GPU rental prices up 106% in 3 months while token prices fall 58% — a16z · 2026-10-31
- vLLM nightly adds PLE disk offload, with a telling [Qwen4Exp] tag on the PR — rmhubbert · 2026-10-31
- Solo AI studio: per-minute GPU rentals rendered a 328-frame Blender shot for ~$1 — r0ck3t23 · 2026-10-31