Question My GPU fans randomly ramping up to %100, requiring a restart.

[Doralkobi]

So for around 3 months now my GPU randomly going crazy, both monitors lose signal and only hard restarting the PC is solving it.
It can happen in any situation, and it seems very random to me.
I did a clean windows installation on a brand new SSD and let it idle for a while to see if it gonna happened again and it did..

Right now I run win 11 (genuine copy) and all drivers are up to date, including bios version.
On idle, the GPU and PSU fans are not working due to very low load until suddenly the GPU gets angry on gods knows what and goes 0-100 in a blink of an eye (PSU is still idling during the GPU rampage).
It can happened in any situation but it happens the most when I just youtubing but it can also happened while gaming or working.

I will try my best to provide as much information on the system as I can:

CPU: I914900K
M/B: Gigabyte Z790 Aorus pro X.
GPU: Gigabyte Aero 4080 super.
RAM: 32GB DDR5 corsair dominator (2X16).
PSU: ROG Loki 850W Platinum (it was the only one available in white!).
SSD: 2TB Samsung 990 pro X 2.
Cooling: Antec 360 water cooling, don't know the full name but it was the only white one available.
Monitors: Main - Gigabyte 32" 1440P 165Hz | Secondary - Dell 32" 4K (runs on 1440) 60Hz.
The room is AC cooled 24\7 due to my server that sits in there.

I use ad pci4 raiser for the GPU due to the fact that its a wall PC, but it is the same one I used with my old 3080TI GPU and it never happened with it.

I am slowly but surely losing my mind here with this one... any help would be very much appreciated.

 

[Fix_that_Glitch]

I would try a new riser. That way you can rule that out. You can try the old troubleshoot of using one stick of ram at a time. Does it do this in safe mode?

 

[Ralston18]

Take a look in Reliability History/Monitor for any error codes, warnings, or even informational events that occur just before or at the time of GPU fans go "0-100".

May not find much with respect to restarts.

However, if gaming or doing other things then you can use Task Manager, Resource Monitor, Performance Monitor, and even Process Explorer to observe what the system is doing or trying to do when the fans spin up.

Use all of the tools but only one tool at a time. It will require that you keep the tool window open and visible to watch. It may also take a bit of trial and error to get the tool window set up in order to see what is happening. Start by focusing on the GPU.

Do as you normally do with added attention when you go into YouTube or other situations that are likely to cause the spin up.

The objective is simply to discover more about the sudden spin ups, the resources being used, to what extent, and the process that is using any given resource at the time of the spin ups.

How old is the build? Especially the PSU.

Has the case been cleaned?

In any case I would power down, unplug, and open the case.

Clean out dust and debris.

Verify by sight and feel that all connections, cards, RAM, jumpers, and case connections are fully and firmly in place.

Also use a bright flashlight to inspect for damages.

 

[Doralkobi]

I will try to use monitoring SW as recommended hoping it will show more info then task manager.
I used TM and nothing spikes before the GPU went bananas but maybe it just lacking information.
I am also open to try a new raiser, any recommendation?

The build itself is around 7 months old
CPU, GPU, PSU, MB, SSD - 7 months
Case, PCI raiser, RAM, cooling - 3 years.
This is a wall open case, there is no dust in it, and everything is accessible and presented.
Picture attached but please ignore the mess of cables, I am working on a solution with 3D models and canals and waiting for DIY cooling parts to arrive to replace the Antec one.

 

[thestryker]

Black screen and 100% fan speed is usually some sort of power issue.

Try setting your PCIe to 3.0 first if you haven't already, but that's the only thing that doesn't require hands on I can think of trying. You can reseat the video card and all power connectors just to ensure everything is securely connected. Otherwise you'd really need to swap the riser cable or plug directly into the motherboard to make sure the riser isn't to blame.

 

[stonecarver]

Okay I have to ask how does GPU react when the side glass panel is removed. Maybe I'm miss something looking at your pictures but every side looks like a closed glass enclosure.

 

[emitfudd]

Okay I have to ask how does GPU react when the side glass panel is removed. Maybe I'm miss something looking at your pictures but every side looks like a closed glass enclosure.

I thought it was a completely open case at first look but I think you are right about the glass. If that's the case, there is zero intake or exhaust that I can see. If the case is wall mounted then the 360mm AIO isn't getting any fresh air either. This is almost comical. Can't wait to see how this plays out.

 

[Doralkobi]

Okay I have to ask how does GPU react when the side glass panel is removed. Maybe I'm miss something looking at your pictures but every side looks like a closed glass enclosure.

Its open at the bottom and the left.
Also there are gaps between every panel.
There is no temperture problems due to the fact that the room is cooled to 20° at all times.
Right now the front panel is removed for safety reasons for around a month and still the GPU crashes.

 

[Doralkobi]

I thought it was a completely open case at first look but I think you are right about the glass. If that's the case, there is zero intake or exhaust that I can see. If the case is wall mounted then the 360mm AIO isn't getting any fresh air either.

The AIO have a filter on the back just as in any case and there is approx. 40cm of space between it and the wall.
This is a very common build and case.
Thermaltake core p5.

 

[thestryker]

Okay I have to ask how does GPU react when the side glass panel is removed. Maybe I'm miss something looking at your pictures but every side looks like a closed glass enclosure.

Bottom and left are open and there should be about an inch between the video card and the side panel. Cooling is certainly not an issue with this case design.

it's one of the Core P5 model cases: https://www.thermaltake.com/core-p5-tempered-glass-snow-edition.html

 

[emitfudd]

The AIO have a filter on the back just as in any case and there is approx. 40cm of space between it and the wall.
This is a very common build and case.
Thermaltake core p5.

40cm is 15". I've never seen any case that is that thick in total, never mind the gap behind the 360mm AIO. I'm confused.

 

[Doralkobi]

Its open at the bottom and the left.
Also there are gaps between every panel.
There is no temperture problems due to the fact that the room is cooled to 20° at all times.

40cm is 15". I've never seen any case that is that thick in total, never mind the gap behind the 360mm AIO. I'm confused.

It has some space from the wall, the mount is a 90" tv wall mount so it add some space there

 

[emitfudd]

The top picture makes it look like it is mounted flush against the wall.

Nobody else has said this and I am mostly saying it in jest, do you have a virus that is allowing a hacker to use your rig for mining? Otherwise I don't know what the issue could be. Hopefully somebody can help you figure it out. If the GPU is only 7 months old have you considered RMA'ing it?

 

[Doralkobi]

The top picture makes it look like it is mounted flush against the wall.

Nobody else has said this and I am mostly saying it in jest, do you have a virus that is allowing a hacker to use your rig for mining? Otherwise I don't know what the issue could be. Hopefully somebody can help you figure it out. If the GPU is only 7 months old have you considered RMA'ing it?

I sure hope I dont haha 😂
I can send the GPU to the lab and they will probebly replace it but I do not think that the GPU is the problem..

 

[stonecarver]

It seems like the sensor that regulates the GPU fans to ramp up the fans here and there is asleep at the wheel.

Can you try MSI afterburner or your chosen program and set a fan curve to 35-45% non gaming and bump up to 60-75% gaming.

If things get stable than it's the card not knowing when to self cool until it's overheating.