MiniMax H3 generates a 5-second video in 363 seconds on AMD Radeon AI PRO R9700
Select_Bug_2844 · reddit · 2026-08-04
- The poster ran MiniMax H3 on an AMD Radeon AI PRO R9700 and generated a 5-second video in 363 seconds.
- They frame that as a strong result for a non-CUDA GPU, since the model is being made to work on AMD hardware rather than NVIDIA.
- The setup used INT8 diffusion checkpoints and an INT4 text encoder, with ComfyUI Portable AMD, ROCm 7.14, and PyTorch 2.12.0+rocm7.14.0 on Windows 11.
- The post is essentially a practical compatibility/performance report for running a video generation model on AMD.
More from Multimodal
- Qwen3.8-Max is claimed to be the best object-detection VLM — iamrobotbear · 2026-08-04
- MiniMax H3 reportedly runs on a Mac, enabling overnight video generation pipelines — AIandDesign · 2026-08-04
- MiniMax H3 LoRA training still fails to preserve character similarity — Kitchen_Carpenter195 · 2026-08-04
- Testing MiniMax H3 in ComfyUI: Joint Video and Audio Generation — GamerVick · 2026-08-04
- How to pass a reference video into MiniMax H3’s native node — RayHell666 · 2026-08-04
- MiniMax H3 lands day-one local serving on 2× RTX 5090s or 1× RTX Pro 6000 — yvbbrjdr · 2026-08-04