Question GPU runs fine for a while until a certain threshold, then loses signal.

[Ackadian]

SO I've had my RX 580 for 4 years at this point (the whole PC is the same age), no signs of deterioration aside from occasional AMD driver crashes, no graphical glitches, no artifacts, none of that usual stuff that points to a GPU dying. A few weeks ago I was playing Fallout 4 and my GPU simply drops signal out of nowhere. I could still hear the game, and was able to hear and talk to the friend I was in a Discord call with, but there's the "no signal" message on the monitor. So I try to restart my PC, and the GPU still refuses to output any signal. After leaving it overnight I managed to get it working again but only for a few minutes. So after a bit of testing, I came to the conclusion that the RX 580 drops signal as soon as it reaches a certain temperature treshold. The 580 comes with 0 RPM settings, so when the GPU is cool as the PC boots up after not being on for a while, everything runs smoothly. But the moment 0 RPM state ends and the fans start spinning, the GPU drops signal instantly.

The weirder thing is that I've tested it in quiet mode, and while it still loses signal (albeit slower due to the slower temperature curve), it is still able to render 3D workloads temporarily without a problem. Tested Cyberpunk and modded Skyrim and there were no graphical glitches of any sort. Framerate is stable and very much the same as it's always been. No problem with loading times either. Adrenaline detects the GPU without a hitch as well. I've already crossed the possibility of it being a PSU issue by switching to a backup GPU I've had laying around, and the same issue does not happen. I have also plugged the 580 into my old PC with a different PSU that I turned into a small testing bench and the signal loss still happens, even after idling in BIOS for a few minutes. I've already tried changing the thermal paste and that did not have any results, though the GPU was not overheating in the first place. With the help of GPT I've deduced that it could be a power delivery issue with one of the capacitors that fails as soon as it tries to draw more power, so I opened it up and looked at all the capacitors to see if any are burnt out or show signs of deterioration and they all look intact. There's no leakage of any sort and no blown tops or discoloration.

So I suppose my question is if anyone has had to deal with something like this. Is it dead for good? Are there any things I could still try to see if it's perhaps a sort of driver issue (doubtful, imo)? It's quite frustrating not being able to at least figure out what happened and I've read that even if it can happen, GPUs don't just die out of the blue like this too often. I can't really afford a new one at this moment, so any help trying to give it some life again is very welcome 🙁

My system specs are:
Gigabyte B450M DS3H Motherboard
AMD Ryzen 5 2600X
16GM RAM
XFX RX 580 8GB OC (faulty one)
RX 560 2GB (backup trash I'm using right now)
Seasonis S12 550W PSU
Windows 10 64-bit
Samsung 870 QVO 1TB SSD

 

[Lutfij]

Welcome to the forums, newcomer!

When posting a thread of troubleshooting nature, it's customary to include your full system's specs. Please list the specs to your build like so:
CPU:
CPU cooler:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
include the age of the PSU apart from it's make and model. BIOS version for your motherboard at this moment of time.

I've already tried changing the thermal paste and that did not have any results, though the GPU was not overheating in the first place.
What thermal paste did you use to replace the one applied from the factory? Did you replace the thermal pads as well?

It's possible that the GPU is on it's way out. The RX580 was a hot headed GPU to begin with.

I've read that even if it can happen, GPUs don't just die out of the blue like this too often.
But they do happen. You could try and reflash the VBIOS on your card and maybe that would help resolve the issue but there's a high chance you might end up with a paperweight if the process goes South.

 

[Ackadian]

Welcome to the forums, newcomer!

When posting a thread of troubleshooting nature, it's customary to include your full system's specs. Please list the specs to your build like so:
CPU:
CPU cooler:
Motherboard:
Ram:
SSD/HDD:
GPU:
PSU:
Chassis:
OS:
Monitor:
include the age of the PSU apart from it's make and model. BIOS version for your motherboard at this moment of time.

I've already tried changing the thermal paste and that did not have any results, though the GPU was not overheating in the first place.
What thermal paste did you use to replace the one applied from the factory? Did you replace the thermal pads as well?

It's possible that the GPU is on it's way out. The RX580 was a hot headed GPU to begin with.

I've read that even if it can happen, GPUs don't just die out of the blue like this too often.
But they do happen. You could try and reflash the VBIOS on your card and maybe that would help resolve the issue but there's a high chance you might end up with a paperweight if the process goes South.

I did not replace the pads, no. But like I said, temperature itself wasn't really the problem. I've been monitoring both temps and voltages on both the PSU and the 580 before the signal losses. The 12V rail never dropped under ~11.930 at worst, and the temperatures only ever reached about 60C degrees when the crash happens.

I also updated the post with all of the system specs I currently know.

 

[Flayed]

You didn't mention what monitor you are using but if you are using a high refresh rate one does the signal loss still happen when you set the refresh rate down to 60Hz?

 

[Ackadian]

You didn't mention what monitor you are using but if you are using a high refresh rate one does the signal loss still happen when you set the refresh rate down to 60Hz?

I'm not sure of the exact model. It's a very old Acer 75Hz TN monitor with one of the long number names
I didn't test this though, but I certainly could. Could you explain why you think this could be related to the issue at all?

 

[Flayed]

I'm not sure of the exact model. It's a very old Acer 75Hz TN monitor with one of the long number names
I didn't test this though, but I certainly could. Could you explain why you think this could be related to the issue at all?

Some of the monitors available have an overclocked refresh rate, from 144Hz to 165Hz for example and its possible for the 165Hz to fail. At least I have heard of one case. I don't think it applies in your case though.

 

[AkroZ]

This can indeed be the monitor getting old if the RX 580 is configured with higher screen size or refresh rate.
But if it is relative to the fan, then most likely the issue is caused by the fan instead of power delivery issue, most likely your fan need more power than designed to start, you should try cleaning it and greasing it: if not brushless you can spin it by hand and see how the break is, like testing the wheel of a bicycle. Evidently use non conductive grease. Often you can found a model number on the back of the fan, you can search online and find the manufacturer selling new ones for some dollars.
On those electrical motors with brushes, those brushes becomes shorter with the friction and stop doing contact. If it is your case, you can change the brushes if you not want to buy another motor (or if you can't find a replacement) but computer fans can be a little fragile and small, you need special tools / techniques to open them without bending the axle.

 

[Flayed]

Are there any things I could still try to see if it's perhaps a sort of driver issue (doubtful, imo)?

You can use Display Driver Uninstaller to remove the drivers and reinstall them if you haven't done so already

Display Driver Uninstaller (DDU) download version 18.1.1.0

Download Display Driver Uninstaller (DDU) Official - Remove AMD/NVIDIA/INTEL Graphics Drivers Cleanly - DDU is a driver removal utility that helps you completely uninstall AMD/NVIDIA graphics card drivers and packages from your system, leaving no leftovers behind, including registry keys...

www.guru3d.com

 

[Ackadian]

This can indeed be the monitor getting old if the RX 580 is configured with higher screen size or refresh rate.
But if it is relative to the fan, then most likely the issue is caused by the fan instead of power delivery issue, most likely your fan need more power than designed to start, you should try cleaning it and greasing it: if not brushless you can spin it by hand and see how the break is, like testing the wheel of a bicycle. Evidently use non conductive grease. Often you can found a model number on the back of the fan, you can search online and find the manufacturer selling new ones for some dollars.
On those electrical motors with brushes, those brushes becomes shorter with the friction and stop doing contact. If it is your case, you can change the brushes if you not want to buy another motor (or if you can't find a replacement) but computer fans can be a little fragile and small, you need special tools / techniques to open them without bending the axle.

But can a faulty fan really cause the whole GPU to just shut down?