This has been driving me mad for the past few days as it doesn't make sense; the longer my PC is under stress, the better it works.
To begin with, I built my rig ~3 months ago, I was very lucky grabbing some rare stuff:
Then work came creeping in and my PC basically turned into a browsing/e-mail machine. About a month ago, my display started randomly turning off or the PC might even reset itself. About half of the times, I could still hear music playing in the background, so it was obviously still running, there was simply no output to my display. I could also sometimes hear the Windows sound when a device gets unplugged.
This was more or less random (after half an hour or it could be 5 hours) and most of the time during trivial things like browsing, writing an e-mail, editing a spreadsheet and so on. Here's the Windows reliability history, you can see it basically appeared overnight. Reliability history report is a mixture of "Windows didn't shut down properly" without any explanation or, more importantly, "Desktop Window Manager stopped working." Here's the crash info if it helps:
I tried doing a clean uninstall of GPU drivers (using Guru Driver Uninstaller), didn't help. I tried 3-month old drivers, back from when I knew it was working, didn't help. I tried various Windows auto repair stuff, nothing.
Here's where it gets weird - if I run GPU stress test like Superposition, the PC resets within a minute or two. It might even reset a second time as soon as it boots back into Windows a few seconds later. However, if I keep doing it, after two or three times, the stress test suddenly stops causing issues and I can repeat it multiple times in a row without my PC resetting. It's as if the PC becomes more reliable once it "warms up."
I also logged GPU stats during the stress test and the temperature never went over 76'C. The GPU clock is hoovering at around 1850MHz. GPU memory temp did reach slightly above 100'C, but I'm reading online that GDDR6X simply runs that hot. I never did any OC whatsoever. I have 2x 110mm fans at the front, another at the back and one at the top, so there should be plenty of airflow.
Just to check, I ran memory test overnight and CPU stress test, everything seems ok.
So I'm basically left with two options; either something with the PSU or GPU. PSU should have enough power with 750W and Gigabyte claims that caps are higher quality from Japan. As for the GPU, I am using the cable splitter that came with the card and each power cable is plugged in it's own PSU port.
I don't know what else to try short of buying a new PSU and re-assembling the entire thing. I'd really like to avoid doing that, so any help, tips or suggestions are highly appreciated. Thank you!
To begin with, I built my rig ~3 months ago, I was very lucky grabbing some rare stuff:
- GPU: NVidia RTX 3080 Founders Edition
- CPU: AMD 5900x
- MOBO: Gigabyte X570 Aorus Pro
- PSU: Gigabyte Aorus 750W 80+ Gold
- RAM: 32GB G.SKILL
- Case: Fractal Meshify C
- OS: Windows 10
Then work came creeping in and my PC basically turned into a browsing/e-mail machine. About a month ago, my display started randomly turning off or the PC might even reset itself. About half of the times, I could still hear music playing in the background, so it was obviously still running, there was simply no output to my display. I could also sometimes hear the Windows sound when a device gets unplugged.
This was more or less random (after half an hour or it could be 5 hours) and most of the time during trivial things like browsing, writing an e-mail, editing a spreadsheet and so on. Here's the Windows reliability history, you can see it basically appeared overnight. Reliability history report is a mixture of "Windows didn't shut down properly" without any explanation or, more importantly, "Desktop Window Manager stopped working." Here's the crash info if it helps:
Code:
Description
Faulting Application Path: C:\Windows\System32\dwm.exe
Problem signature
Problem Event Name: APPCRASH
Application Name: dwm.exe
Application Version: 10.0.19041.746
Application Timestamp: 6be51595
Fault Module Name: KERNELBASE.dll
Fault Module Version: 10.0.19041.804
Fault Module Timestamp: 0e9c5eae
Exception Code: e0464645
Exception Offset: 000000000010bd5c
OS Version: 10.0.19042.2.0.0.256.121
Locale ID: 8192
Additional Information 1: 0503
Additional Information 2: 05038a8d0f1df7d435a57df411af2cff
Additional Information 3: e54a
Additional Information 4: e54abd716d1fac5ba98f2157a857cfe1
Extra information about the problem
Bucket ID: 623e25242da0256adf46691f2c2601cb (2253604246151102923)I tried doing a clean uninstall of GPU drivers (using Guru Driver Uninstaller), didn't help. I tried 3-month old drivers, back from when I knew it was working, didn't help. I tried various Windows auto repair stuff, nothing.
Here's where it gets weird - if I run GPU stress test like Superposition, the PC resets within a minute or two. It might even reset a second time as soon as it boots back into Windows a few seconds later. However, if I keep doing it, after two or three times, the stress test suddenly stops causing issues and I can repeat it multiple times in a row without my PC resetting. It's as if the PC becomes more reliable once it "warms up."
I also logged GPU stats during the stress test and the temperature never went over 76'C. The GPU clock is hoovering at around 1850MHz. GPU memory temp did reach slightly above 100'C, but I'm reading online that GDDR6X simply runs that hot. I never did any OC whatsoever. I have 2x 110mm fans at the front, another at the back and one at the top, so there should be plenty of airflow.
Just to check, I ran memory test overnight and CPU stress test, everything seems ok.
So I'm basically left with two options; either something with the PSU or GPU. PSU should have enough power with 750W and Gigabyte claims that caps are higher quality from Japan. As for the GPU, I am using the cable splitter that came with the card and each power cable is plugged in it's own PSU port.
I don't know what else to try short of buying a new PSU and re-assembling the entire thing. I'd really like to avoid doing that, so any help, tips or suggestions are highly appreciated. Thank you!
