MiniMax H3 I2V on a 5090: cu130 torch fix kills OOM, 0.98MP x 20s in ~21 minutes
Realistic-Fennel-190 · reddit · 2026-11-14
A detailed Reddit writeup of running MiniMax H3 image-to-video on a single RTX 5090 (32GB) with ComfyUI 0.45.0:
- The problem: a cu128 torch build disables comfykitchen's CUDA backend on ComfyUI 0.45.0, falling back to eager mode; 0.8MP x 20s with turbo 8 steps OOM'd at step 0 in int8linear (23.9GiB allocated, 6.24GiB more requested).
- The fix: one line—pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu130. After restart, torch 2.14.1+cu130 enables the CUDA backend and the same workflow runs with no OOM.
- Benchmarks (20s, turbo 8 steps, 9:16, same seed): 0.8MP → 13:38, 0.9MP → 17:43, 0.98MP → 20:51 (20:06 of that sampling); at 0.98MP VRAM sat at 31824/32607 MiB, 566W, 74°C.
- Prompting tip: reusing a 15s prompt for 20s caused drift in the last 5 seconds; writing four 5s phases and keeping wind away from the face fixed it.
Stack includes the minimaxh3 int8 unet, qwen3vl 32B text encoder, and lightx2v turbo 8-step LoRA. Author is asking whether 24GB cards can handle it.
More from Infra
- Openship: self-hosted deployment platform with built-in CI/CD hits 15.3k GitHub stars — jackedAJ · 2026-11-14
- Run Gemma 4 locally with one Ollama command: 7-10GB, offline, no subscription caps — nikola_mr64990 · 2026-11-14
- Inkling Small 2 fits on 4x RTX Pro 6000 with 7.5M token KV cache at 120 tok/s — LiTianleli · 2026-11-14
- Perplexity team teases minimalist LLM trainer built with coding agents, open source soon — BanghuaZ · 2026-11-14
- Serving many fine-tuned variants of a 27B model on one GPU: 4-bit CQ format and 55x smaller adapters — textclf · 2026-11-14
- Musk: Future Starship Flights Will Carry Over 50 Tb/s of Starlink Bandwidth to Orbit — elonmusk · 2026-11-14