Running Krea2 on RTX 5090: ComfyUI Setup Faces VRAM Bottlenecks
orangeflyingmonkey_ · reddit · 2026-07-31
A user running Krea2 locally in ComfyUI on an RTX 5090 (32GB) reports hitting VRAM limits. The UNet and Qwen3VL text encoder combined demand 32.9GB, preventing both from staying resident simultaneously. This forces ComfyUI to reload the text encoder from disk for every generation, severely slowing down the workflow. The user is seeking recommendations for a lighter text encoder, setup tweaks to avoid reloads, and other VRAM optimization tricks specific to Krea2.
More from Infra
- SK Hynix Hits 76% Margin on HBM: Are Markets Mispricing AI Infrastructure? — km · 2026-07-31
- Cloudflare AI Search Launches Native Integration for LangChain and AI SDK — irvinebroque · 2026-07-31
- Viewpoint: Low Interest Rates Indicate We Are Not Overinvesting in AI Compute — tszzl · 2026-07-31
- OpenAI Slashes GPT-5.6 Prices by Up to 80%, Undercutting Rivals — Simon Willison · 2026-07-31
- Optimizing LLM Inference TPS on NVIDIA Blackwell Is the Funnest Thing — abhijithneil · 2026-07-31
- Cloud Giants Scale Up: Google Cloud Revenue Surges 82% YoY — davidyin44 · 2026-07-31