Integrated Video on Motherboard #14
Replies: 2 comments
-
That would be cool. Hope to get more info on this. |
Beta Was this translation helpful? Give feedback.
-
Hey, @GKartist75 and @XobyThePoet - I haven't been utilizing the "Discussions" part of the repo, so if it seems like I have been ignoring the discussion, it is because I have! ;) So, Comfy has code in its model_management.py for intel integrated GPUs ( try:
import intel_extension_for_pytorch as ipex
_ = torch.xpu.device_count()
xpu_available = xpu_available or torch.xpu.is_available()
except:
xpu_available = xpu_available or (hasattr(torch, "xpu") and torch.xpu.is_available())
try:
if torch.backends.mps.is_available():
cpu_state = CPUState.MPS
import torch.mps
except:
pass Have you tried downloading the drivers and seeing if Comfy recognizes the device? At the very least, the VRAM should be available to you for storage of model components, opening up more for latent operations. Also - I would be remiss if I didn't mention that you already have another Torch-compatible device in your system - your CPU/DRAM. You can already offload CLIP and VAE there should you wish! |
Beta Was this translation helpful? Give feedback.
-
is it possible to select the integrated video on a motherboard to boost some extra performance (e.g, run a clip node?), or will this be to slow or has a negative impact?
Beta Was this translation helpful? Give feedback.
All reactions