InvokeAI
InvokeAI copied to clipboard
invoke-ai
[bug:] PC shuts down during an iteration
Is there an existing issue for this?
- [X] I have searched the existing issues
OS
Windows
GPU
cuda
VRAM
4GB
What happened?
During an outpainting iteration, pc just shutdown (hard), i suspect its temperature issue, but it never happened before prior with InvokeAI.
I tried to sift in the Windows Event Manager or any memory dumps for as to why, but none i could find.
I have updated to InvokeAI to 2.2.0
Screenshots
No response
Additional context
No response
Contact Details
No response
Make sure your power supply has enough power at peak for your video card. If you have ever upgraded your video card but not your power supply, it could be that your video card tried to draw more power than your power supply could handle, thus causing a hard shut down mid-iteration (when the video card is running at 100%). Power failures don't leave behind traces in the windows logs typically, which is why I'm suspecting this is your issue.
Make sure your power supply has enough power at peak for your video card. If you have ever upgraded your video card but not your power supply, it could be that your video card tried to draw more power than your power supply could handle, thus causing a hard shut down mid-iteration (when the video card is running at 100%). Power failures don't leave behind traces in the windows logs typically, which is why I'm suspecting this is your issue.
I got a 650w power supply, using a lightweight nvida 1050ti, could this card really pull more then that ?
Make sure your power supply has enough power at peak for your video card. If you have ever upgraded your video card but not your power supply, it could be that your video card tried to draw more power than your power supply could handle, thus causing a hard shut down mid-iteration (when the video card is running at 100%). Power failures don't leave behind traces in the windows logs typically, which is why I'm suspecting this is your issue.
I got a 650w power supply, using a lightweight nvida 1050ti, could this card really pull more then that ?
For a healthy power supply, no it shouldn’t. But when your system just cuts off in the middle of 100% load (which is what happens when it’s running steps), 9 times out of 10 it’s either an underpowered power supply, or a dying one.
Well thanks Diego, i guess i'm off buying a new power unit :P
you can run a stress test to identify if its 100% for sure power supply related or thermals, given that its a 10 series and assuming you use it daily i would think the thermal paste within has questionable qulity as of 2022. i would download furmark and find a way to look at thermals, both for cpu and gpu, keep a close watch,,, i reccomend furmark and run it at whatever settings stress out your card without going overboard
1050ti isnt too power hungry for the 10 series cards, if you baught your pc from a store they would have accounted for 100% load, if you built it yourself then its something you should have accounted for.
of coarse give your pc the best possible chances, make sure you have fans and fins dusted, if it is thermals and you have sufficient gonads i recommend liquid metal (gallium) as long as your heat sinks are copper NOT ALUMINUM and you know this 100%. gallium thermal transfer is basically like welding your die to the heatspreader. you can always use fresh silicone paste.
i did liquid metal for my laptop and havent had any crashing issues and cpu stays 4.0ghz 85* max gpu 75-85*max (+200mhz clock offset...) depending on room temp, and i leave this thing running all night fine tuning models...
i also got a 10 series and at least for a laptop thermals have been a larger issue. the only time power was an issue was on battery but it should crash as soon as you click start rendering... once it hits 100% load not 30 seconds into the render thats a clear sign the thermal mass of your gpu was saturated and wasnt able to cool off... if it crashes almost instantly then power demand wasnt met
I already bought a 750w unit, just in case, and i did build this system myself. I do believe its the peak power that poked the PSU a bit hard.