[SOLVED] GPU Randomly Spiking to 100%

[dominicwild]

This has happened a few times randomly with my GTX 2080 Ti. It spikes to 100% GPU usages, when nothing is going on and Windows 7 turns off its aero effect as a result.

I've tried using process explorer to see what would be using it up, however when I opened up process explorer to see, it didn't show any process using the GPU. At least, nothing that added up to almost 100% usage.

These are some of the graphs MSI had following the spike. If they'd be of any help.

This strange behaviour follows many other things after my graphics card change, from slow boots and lots of random blue screens.

Any help would be appreciated.

 

[Phaaze88]

Post full PC specs.
Also, you do notice that you have cpu threads running 90C+, right?

 

[dominicwild]

Post full PC specs.
Also, you do notice that you have cpu threads running 90C+, right?

Intel Core i7 4790K
CPU Cooler: Corsair H100i Extreme Water Cooler
OS: Windows 7 PRO
Motherboard: Asus Maximus VII Ranger
RAM: Corsair CMZ32GX3M4X1600C10 Vengeance 32 GB (4 x 8 GB) DDR3 1600 Mhz C9 XMP Performance Memory
Hard Drive: 1TB SSD Sandisk
Secondary Hard Drive: Seagate 1TB (Hybrid 8GB SSD)
Third Hard Drive: Kingston 120GB SSD
Graphics card: NVIDIA GeForce RTX 2080 Ti 11GB
PSU: 750W Corsair RM Fully Modular

Also, I do notice that. The CPU's spiked when the GPU did. Although CPU usage didn't.

 

[Phaaze88]

CPU Cooler: Corsair H100i Extreme Water Cooler

You've got a more immediate problem via the cpu cooler. It's either:
-clogged: it will happen over time, so nothing to really do here but keep the fans and toss the rest out
-pump failure: same as above
-not configured: the pump should be set to PWM mode in bios, then set to run at 100%.

In the case of the first 2, I hope you have a spare air, or other liquid cooler, on hand...
You can check by feeling the cpu block, tubing, and the radiator: they should all feel fairly warm. If the heat is focused in the cpu block, then there's a problem.

 

[dominicwild]

You've got a more immediate problem via the cpu cooler. It's either:
-clogged: it will happen over time, so nothing to really do here but keep the fans and toss the rest out
-pump failure: same as above
-not configured: the pump should be set to PWM mode in bios, then set to run at 100%.

In the case of the first 2, I hope you have a spare air, or other liquid cooler, on hand...
You can check by feeling the cpu block, tubing, and the radiator: they should all feel fairly warm. If the heat is focused in the cpu block, then there's a problem.

Looking at it, I opened the case and felt what you said.

https://imgur.com/WKZeJfL

View: https://i.imgur.com/WKZeJfL.png


That is what I felt. All were warm, nothing too extreme in terms of heat. However the bar at the heatsink (?) at the back was noticeably hotter than the rest. So sounds like it should be working?

I checked the status with some software called Corsair link, as well and that seems to indicate it is functioning normally.

Here are some more temperatures from the Corsair LINK program. Seems to say quite a few.

https://imgur.com/D3tY0pH

View: https://i.imgur.com/D3tY0pH.png


It has a temperature of 127 on the mother board? However this keeps shooting from 0 to 127.

https://imgur.com/ADzcSut

View: https://i.imgur.com/ADzcSut.png


So may just be a faulty sensor?

Looking at when I play games and do something actually intensive, my CPU walks up to 90-100 degrees. Temperature across my four cores is generally:

https://imgur.com/KAtyw0c

View: https://i.imgur.com/KAtyw0c.png


When playing a game. Must have been the case for quite some time now, considering how long this problem has been going on for.

 

[Phaaze88]

That 'bar' is one of the motherboard's VRM heatsinks.
One of the cons of liquid coolers is that the VRMs don't get the direct airflow that's usually provided by air coolers; they'll naturally run warmer as a result.

Sounds like the unit is running normally; if it weren't the cpu block would've been hot and everything else quite cool.
Sometimes the software trips up, so the above is a sure-fire way to check the cooler.

The motherboard has several thermal sensors - I don't even know where half of 'em are located, other than some key areas, like the chipset, I/O control, and VRMs.

That first Corsair Link image: those are idle core temps? That's high if it is.
Plus, in the very first image, the core frequencies were observed dropping way down near the end...

 

[dominicwild]

That first Corsair Link image: those are idle core temps? That's high if it is.
Plus, in the very first image, the core frequencies were observed dropping way down near the end...

Yep, those are idle temperatures. They're typically all around 70 or mid 60's. It is strange the core clocks jumped around there, as most of the time they're stable 4398.

But that was during that abnormal GPU spike.

 

[Phaaze88]

That's not good.
Max out the pump speed if it isn't already, and check again. If it already is, there's either a clog somewhere, or air pockets getting into the pump.

I also missed the 38.4C liquid temp - that's warm too.

The gpu spikes are likely some 3rd party software doing it - might just be Corsair Link doing it, but I've never used it, so I'm not familiar with it.

 

[dominicwild]

That's not good.
Max out the pump speed if it isn't already, and check again. If it already is, there's either a clog somewhere, or air pockets getting into the pump.

The gpu spikes are likely some 3rd party software doing it - might just be Corsair Link doing it, but I've never used it, so I'm not familiar with it.

Yeah, I've maxed out everything possible when it comes to fans and such.

I don't think some software is causing the spikes actually. They haven't happened frequently. When it happened, I opened process explorer and it showed no application actually using GPU, despite it maxing at 100% usage. Only things like chrome were using about 5-10%.

There are lots of other abnormalities that have happened since replacing my old card. Windows stays on a black screen for a good minute or three every time it boots now. Before everything boots up. I can force close and open explorer.exe through Task Manager though to speed up the process to my desktop.

 

[Phaaze88]

It could use some cleaning. Remove the fans and blow the dust out of the radiator - heck, may as well clean the rest of the PC while you're at it.
Remove the old paste and apply new, etc.

Did you use DDU(Display Driver Uninstaller) to remove the old gpu drivers? There must be something using it, however brief.
Gpu Scheduling got ruled out because you're still on Windows 7.

 

[dominicwild]

I'll try to clean it out and see what happens. Can dust increase temperature dramatically? To the point of 20-30 extra degrees? Assuming normal temperatures are around 30-40.

I used DDU to uninstall the drivers before and then install the new ones.

Yep, still on Windows 7, so wouldn't be that.

 

[Phaaze88]

If you haven't done it in like a year or more, yeah.
I was going to get around to it eventually, as it was pretty visible in the 2nd image. Dust in the fans, trapped in the rad... less air going in and getting out.

 

[dominicwild]

If you haven't done it in like a year or more, yeah.
I was going to get around to it eventually, as it was pretty visible in the 2nd image. Dust in the fans, trapped in the rad... less air going in and getting out.

I had the water cooler for a total of around 6 years. I just replaced it. I'm now getting temperatures around 50 degrees, high 40's. Which sounds about right, given the ambient temperature under my desk is around 31 degrees. Under load, it doesn't get anywhere near 100 anymore. The max it got to was 71.

I think the water coolers coolant evaporated or something? As I shook it and hardly heard any water. When I shook this new one, I heard tonnes more. It was shaking an almost empty bottle, vs half full bottle.

So must have just been the water cooler wearing out.

Thanks for the help.

 

[Phaaze88]

Permeation is taking place over time. Nothing you can really do about it, as that unit wasn't intended to be refilled.

That cooler put in some good time; it was time for it to go.