Seemingly radnom restarts

[Neqs20]

Some context info: electric installation in my home is pretty new (house was built 6 years ago), but sometimes there are some weird interactions with appliances. Turning off the light near the TV sometimes makes it show static on the bottom half of the screen. A week ago there was a short power loss (lasted less than a second), but it was enough for my PC to restart. Loaded up the system again and everything was fine.

 

3 days ago my PC randomly crashed for the first time. It happened on Fedora while playing a game. There were no errors shown, just a sudden black screen for a few seconds followed by a boot sequence. The crash also happens on Windows 10, with no BSOD. I immediately went to inspect the PC to find the cause, but to no avail. Tried looking for burned regions on MoBo and found nothing. Changed thermal paste, unplugged unnecessary cables (front IO, etc.) but it still crashed.

 

After multiple attempts I was able to reach login screen on my Fedora, switched to tty and tested memory with memtester (couldn't make memtest86 to work). All tests passed. This was the time I realized it doesn't crash in CLI mode, only in GUI.

 

I found that the crash may occur in the following cases (both on Fedora and Windows):
    * after seeing 1 frame of the login screen it automatically crashes
    * after typing in my password and pressing enter
    * after toggling calendar or notifications list
    * after opening certain software (discord, firefox)
    * after pressing windows key
    * some other unexplained random events

 

I tried to reproduce the crash with a Radeon R7 340/240 GPU I found in my old PC and to my surprise it never crashed on either systems.

 

Now that my system was stable I collected logs from crashes on both systems and followed solutions from multiple threads, which include:
    * Revert default BIOS settings and update BIOS.
    * Test single RAM stick in the first socket
    * Update chipset
    * Disabling c6 state in fedora
    * Disabling Global C-state Control in BIOS
    * Changing Power Supply Idle Control to "Typical Current Idle" in BIOS
    * Plugging GPU to the other PCIe slot.
    

None of these helped. PC will always reboot at some random point if 5500XT is installed.

 

Clean installing Windows and then Fedora on the same drive didn't fix the issue either. I was also having a hard time booting into the installation media, so this rules out problem with the drives.

 

Windows' built-in memory diagnostic tool also did not find any problems with memory.

 

Throughout all the threads I read I noticed that there was 1 thing in common: CPU. People seem to have problems with 3rd gen Ryzens, especially on full AMD configurations with Gigabyte boards.

At this point I am out of ideas. With all this information it's hard for me to tell which part is faulty, is it the PSU, CPU, GPU, RAM or sth else?

 

PS. Sorry if some parts are unclear. I can explain further if needed.

Specs (build it around 3 years ago):

CPU: Ryzen 9 3900X
PSU: Seasonic Focus PX-650 80+ Platinum
GPU: Radeon RX 5500XT
MB: Gigabyte B550 GAMING X V2
Memory: G.Skill DDR4-3200 2x8 GB
Drive: SSD 860 QVO 1TB (Windows)
Drive: NVMe 970 EVO Plus 1TB (Fedora)

 

windows_event_viewer_error5.txt windows_event_viewer_error4.txt windows_event_viewer_error3.txt windows_event_viewer_error2.txt windows_event_viewer_error1.txt fedora_mcelog.txt fedora_journalctl.txt fedora_dmesg_grep_mce.txt

[Neqs20]

1 hour ago, TimedPing said:

I suspected that you have bad electrical installation and it's affect your PC slowly (especially PSUs and motherboard Power Delivery system). Something I can suggest is bring your PC to other location with known good power source and run full tilt benchmark to see the anomaly, if it's still happened, change PSUs and Motherboard temporarily and pinpoint which hardware that have the annomaly.

Thanks, I will be contacting an old friend of mine that built my old PC. He might have some spare parts to help me diagnose this.