Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

KSampler slowing down when processing is interrupted #5471

Closed
traugdor opened this issue Nov 2, 2024 · 8 comments
Closed

KSampler slowing down when processing is interrupted #5471

traugdor opened this issue Nov 2, 2024 · 8 comments
Labels
AMD Issue related to AMD driver support. User Support A user needs help with something, probably not a bug.

Comments

@traugdor
Copy link

traugdor commented Nov 2, 2024

Your question

image

screenshot of my console. I was interrupting it because I could tell from the preview that it wasn't going to be the image I wanted and I have a pretty large workflow that would take a long time to finish processing.

It started out pretty fast (for my hardware) and then got slower with each interruption.

I haven't noted this behavior until just recently. I think it has something to do with interrupting the workflow.

Logs

No response

Other

OS: Windows 10 Professional
Hardware: CPU Ryzen 7 3700x, GPU Powercolor Radeon 6600xt

Not sure what else is needed. Happy to provide information to help diagnose this issue.

@traugdor traugdor added the User Support A user needs help with something, probably not a bug. label Nov 2, 2024
@zeng121
Copy link

zeng121 commented Nov 19, 2024

I had the same problem

@ltdrdata
Copy link
Collaborator

Is there no performance degradation if you let the workflow complete fully without interrupting it?

@traugdor
Copy link
Author

traugdor commented Nov 19, 2024

Is there no performance degradation if you let the workflow complete fully without interrupting it?

not on my end, no, not consistently. Only when interrupting the workflow and starting it again is it consistently reproduce-able. Could be a caching issue? I'm running pretty close to my max VRAM. I'm hesitant to download the latest of ComfyUI and try again simply because I have made edits to some files to fix issues found in #5427 and #3539 pertaining to AMD cards running under DirectML.

when I let it run to completion, the run times of the workflows seems to be most consistent with only slight variations here and there.

@traugdor
Copy link
Author

When letting it run to completion it seems that every other run will be significantly faster than the one before
image

@ltdrdata
Copy link
Collaborator

When letting it run to completion it seems that every other run will be significantly faster than the one before image

After interrupting, try performing the next execution after taking a sufficiently long break.

@traugdor
Copy link
Author

It seems to be exhibiting this behavior more consistently when I am using multiple monitors.

I think it's literally just a VRAM issue at this point.

@huchenlei huchenlei added the AMD Issue related to AMD driver support. label Dec 16, 2024
@traugdor
Copy link
Author

traugdor commented Jan 2, 2025

Just updating to say that the issue has gotten worse and now I can barely generate a 512x512 image more than a few times before getting allocation errors. Is a VRAM cache not clearing properly? Stopping the Comfy process in Powershell and restarting it usually fixes it, but I cannot generate more than 2 sometimes 3 images without having to restart.

@traugdor
Copy link
Author

traugdor commented Jan 2, 2025

I downgraded the code several time to a copy from October and the issue still persists, so I think I have a hardware issue.

@traugdor traugdor closed this as completed Jan 2, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
AMD Issue related to AMD driver support. User Support A user needs help with something, probably not a bug.
Projects
None yet
Development

No branches or pull requests

4 participants