• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

GPU crashing with artifacts under load

N

NoName0815

New Member
Joined
Jan 3, 2020
Messages
10 (0.01/day)
I recently swapped the cooler on my gigabyte r9 390 because the stock one is just too loud.
After installing the gpu, i tried running furmark as a test if everything works fine, but my pc immediatly crashed and i could only turn it on again after switching the power supply off and on again. Interestingly, the power LED and the RGB light of my ram were still on and my keyboard was flickering in every color possible, but my CPU cooler (also RGB) had no light on at all.
After I tried to play a game, my PC crashed "slower", or in other words I was able to see that some weird artifacts are happening (white big pixels going diagonal from top left to bottom right)
I tried to find a solution by myself but nothing worked.

My System:
Ryzen 5 2600 (stock)
Gigabyte B450 Aorus M
Ballistix Tactical Tracer (2666 variant, oc 3000)
Gigabyte r9 390 (stock, with Prolimatech MK-26 cooler)
be quiet! Pure Power 10 600w
 
Furmark is the worst software to run on a GPU as it puts a unnecessary load on the GPU. Most manufacturers advise against using it.

Likely issue is the cooler not properly mounted and the GPU is overheating badly causing hardlocks

Did you have this issue before swapping the cooler
 
I did not have this issue, but i checked the temps and they stayed und 50°C.
As I said in the original post, I also tested with a game.
 
Remount that cooler...

...what are temps of the card when gaming? 50C load doesnt seem right at all....for that 250W card...
 
Are you certain the cooler has made contact with the actual GPU?

Remove it, remove the paste, and replace it, then remount. Also, how is the cooler cooling the VRAM? Is it making contact like the one from the factory likely did?

Finally, please don’t use Furmark after you get this sorted.
 
  • Like
Reactions: 64K
Another issue i am finding is that with a simple search the cooler you listed doesn't list compatibility with the R9 390.

I also see that cooler doesn't include fans. Did you mount fans to it.


And if the issue wasn't present before cooler swapping then the issue has been caused by the cooler change.
 
Artifacting is usually a VRAM issue. My guess without more info is that your VRAM is overheating.
 
The cooler does fit on an r9 290, which is basically the same but less vram
The mounting holes actually were pretty perfect if you ask me, and sadly the card doesn't let me check vram or vrm temps

But i also tested th vram with occt, it didn't crash

Durvelle27 said:
Another issue i am finding is that with a simple search the cooler you listed doesn't list compatibility with the R9 390.

I also see that cooler doesn't include fans. Did you mount fans to it.


And if the issue wasn't present before cooler swapping then the issue has been caused by the cooler change.
Click to expand...
yes i did, twi noctua industrialppc 3000 fans
 
NoName0815 said:
The cooler does fit on an r9 290, which is basically the same but less vram
The mounting holes actually were pretty perfect if you ask me, and sadly the card doesn't let me check vram or vrm temps

But i also tested th vram with occt, it didn't crash
Click to expand...
You must realize you also have a aftermarket GPU which are not based off reference PCBs so there is a difference.
 
As far as i know, gigabyte used the same mounting holes like the reference card
 
Did you mount heatsinks to the VRAMs

As it has very hot VRAM chips that need to be properly cooled
 
What kind of cooling or contact does that cooler make with the VRAM?
 
64K said:
Artifacting is usually a VRAM issue. My guess without more info is that your VRAM is overheating.
Click to expand...
Most gddr6 isnt cooled and works fine without...

Remount...

Edit: which specific gigabyte r9 390 do you have? Product number or link, please.

NoName0815 said:
As far as i know, gigabyte used the same mounting holes like the reference card
Click to expand...
mounting holes are typically the same, that doesnt mean the cooler fits properly....
 
EarthDog said:
Most gddr6 isnt cooled and works fine without...

Remount...
Click to expand...
as i said, I did the occt vram test and it didn't crash. If that test isn't realistic (in terms of heat) just tell me and i would test in different ways
 
EarthDog said:
Gddr5 didnt really need it either. Most memory doesnt at stock...
Click to expand...
But from my own personal experience Hawaii was a very hot GPU that used GDDR5 at high high volts to sustain clocks in turn causing the VRAMs to run very hot. Almost all AIBs mounted a heat plate over them to combat this but even than they still ran hot.
 
EarthDog said:
Most gddr6 isnt cooled and works fine without...
Click to expand...
It’s an R9 390, so GDDR5, I believe. I remember those 380/380X/390 cards ran very warm all around and had high VRAM temps in reviews.
 
Durvelle27 said:
But from my own personal experience Hawaii was a very hot GPU that used GDDR5 at high high volts to sustain clocks in turn causing the VRAMs to run very hot. Almost all AIBs mounted a heat plate over them to combat this but even than they still ran hot.
Click to expand...
Exceptions arent rules. ;)

Gddr5/gddr6... my statement doesnt change, boys. :)
 
Durvelle27 said:
But from my own personal experience Hawaii was a very hot GPU that used GDDR5 at high high volts to sustain clocks in turn causing the VRAMs to run very hot. Almost all AIBs mounted a heat plate over them to combat this but even than they still ran hot.
Click to expand...
since you are so sure that the vram is overheating, is that occt vram test not realistic?
 
Alright OP this is my suggestion

Take the cooler off and reapply TIM then reinstall the stock cooler. The new TIM will help with thermals. Once installed apply a custom fan curve to combat noise.

Run a loop of uniengine heaven and monitor temps

HWinfo64 will give you temps for VRAMs and Core
 
Durvelle27 said:
Alright OP this is my suggestion

Take the cooler off and reapply TIM then reinstall the stock cooler. The new TIM will help with thermals. Once installed apply a custom fan curve to combat noise.

Run a loop of uniengine heaven and monitor temps

HWinfo64 will give you temps for VRAMs and Core
Click to expand...
alright will try, but why doesn't gpu Z show vram if HWinfo64 can do so?
 
Durvelle27 said:
Furmark is the worst software to run on a GPU as it puts a unnecessary load on the GPU.
Click to expand...
I agree Furmark puts an unnecessary load on the GPU, but I note that is the exact the purpose of stress tests.

However, it is like taking your car to the drag strip to see how fast it can go in a 1/4 mile. It is extremely abusive which is why warranties won't cover repairs when it blows up.

Plus, it is a "synthetic" scenario. You would not drive it like that out in the streets. So I also agree, don't do it. Test your card by using it performing the tasks you bought it for.

In the meantime, I agree with rtwjunkie - remove the cooler, thoroughly clean the mating surfaces, and reapply a fresh, new, as thin but complete layer of new TIM. Then remount the cooler.

Run the system through a couple of heat up/cool down cycles to allow for TIM curing time. Make sure your case is providing a decent supply of cool air flowing through the case, then use the computer normally and see how it does.
 
You must log in or register to reply here.
Back
Top