How to get ComfyUI running on your new Nvidia 5090 or 5080.
ComfyUI has day 1 support for the new 50 series Blackwell GPUs.
Thanks to Nvidia ComfyUI has support for the new 50 series GPUs on launch day.
If you managed to get your hands on one you can see this discussion thread for how to get started. Be sure to check it often as I will keep it updated as much as possible and I expect there will be many performance improvements and updates over the next few months.
You need a pytorch build with CUDA 12.8 or newer, anything lower than that will not work. For how to get that see: the discussion thread
For Windows users still struggling with Blackwell setup — I packaged a one-click Windows-native solution that uses cu130 (for full NVFP4 support) with 28 verified custom nodes:
https://github.com/hiroki-abe-58/ComfyUI-Win-Blackwell
No WSL2, no Docker. Just `setup.bat` and wait ~20 minutes.
I'm using the sytem like below:
pytorch version: 2.5.1+cu124
xformers version: 0.0.28.post3
Device: cuda:0 NVIDIA GeForce RTX 4080
Python version: 3.11.9
ComfyUI version: 0.3.60
ComfyUI frontend version: 1.26.13
Blow errors came out while running the WanAnimate worklow based clip model umt5-xxl-enc-bf16 (umt5-xxl-fp8-e4m3fn_scaled was ok but the video quality was bad).
#### It seems that models and clips are mixed and interconnected between SDXL Base, SDXL Refiner, SD1.x, and SD2.x. Please verify. ####