Question GPU not outputting video until CMOS reset is pressed

[wownickc]

So this is a weird one, recently I bumped my computer with my knee under the desk, the picture froze (audio still going fine), and I turned off the computer by holding the power button and turned it back on.

After doing that, I got no output, had a little freak-out then got to troubleshooting. I plugged an HDMI cable into my motherboard to use integrated graphics, and device manager showed no sign of my GPU, so I had another freak-out.
Then I got troubleshooting again and pressed the CMOS button on the back of my mobo and saw that my gpu got recognised and outputted video, yay. But this was only temporary, it happened again after playing some satisfactory, a pretty GPU intensive game.
So I repeated the process, which involved turning the computer off, unplugging power, pressing the CMOS button, and rebooting. I checked in the bios for things like default video output, anything that sounded like it would mistakenly switch outputs etc.
I then updated my bios in the hopes it was a bios related issue, and pressed the 'Load Optimized Defaults button', and did a couple reboots and saw that the issue didn't happen again, so I think my problem is solved right? wrong.
Played some satisfactory, for a bit to get the GPU warmed up (water-cooled with aio so only reached about 56 degrees C), did a reboot test and it happened again. Turned off the PC and turned it back on again, this time without a CMOS reset, and it booted up fine.
Could be something related to GPU temp? could be something Bios related? Maybe even a freakish satisfactory bug? I don't know.
There are so many different troubleshooting strands I haven't tied up I don't know where to go but here!

A few notes:
I was playing satisfactory when the issue happened.
I have had no GPU issues before this.
I haven't checked the bios when it switches to internal graphics.
I have reseated the graphics card with no change.
After the first successful boot following the problem, satisfactory had seemingly partially uninstalled itself from steam.

Edit: played different game for prolonged period and a reboot led to the problem once again. Leading me to suspect the motherboard doesn't recognize the gpu because of temperature or something?

Specs:
i9 9900k - aio water cooled
RTX 2080ti - aio water cooled (kraken g12)
Corsair vengeance LPX 32gb Ram
Seasonic Prime TX - 850W
Samsung 960 evo NVME 1tb
NZXT H710
Windows 10

 

[Lutfij]

Welcome to the forums, newcomer!

Make and model of your motherboard? Which version of Windows 10 are you currently on? We're on version 20H2. See if the issue you're facing is seen when you breadboard the system with the bare minimum(outside the case), perhaps you're suffering from a minor short in your system?

^ Considering this all began after the bump by the knee to the system.

Fire up Samsung's Magician app and see if your SSD has any firmware updates. You might also want to install the NVMe drivers for your Samsung SSD, it's found on the support site on Samsung.

 

[wownickc]

Welcome to the forums, newcomer!

Make and model of your motherboard? Which version of Windows 10 are you currently on? We're on version 20H2. See if the issue you're facing is seen when you breadboard the system with the bare minimum(outside the case), perhaps you're suffering from a minor short in your system?

^ Considering this all began after the bump by the knee to the system.

Fire up Samsung's Magician app and see if your SSD has any firmware updates. You might also want to install the NVMe drivers for your Samsung SSD, it's found on the support site on Samsung.

Thanks for the welcome!
Sorry forgot to put the motherboard, its a gigabyte Aorus Master z390 (not the G2 version), so its got bells and whistles for diagnosis. I may have found a solution, although it doesn't explain the initial happening:

After resetting CMOS, my fan curves got reset. The Kraken G12 places an aio water block on the gpu core, and provides a fan to cool the VRM on the GPU, this fan ideally goes at full tilt the whole time.
However, having its curve reset gave it a generic temperature curve linked to the temperature of the gpu core, which obviously is water cooled. This meant that, while I played games, the GPU core was well cooled, but the fan cooling the vrms thought it didn't need to do much, running at around 400 rpm compared to its usual 1400. I didn't experience any artifacting or stuff while playing, but cranking the curve back up allowed me to shut down and boot up right after playing a game.

So it seems that it was the GPU VRMs preventing my motherboard from recognising the GPU until they had cooled.

I have noticed previously that a bump can cause the PC to reboot, so I assume that is some kind of short that I will try and find by fiddling around with the motherboard standoffs and if that doesn't work then bread boxing. Does a reboot imply a short between the motherboard and the case?

In reply to your questions:
I haven't yet updated to 20H2, probably a good idea though so thanks for reminding me.