GPU artifacts, need help

[Daraeon]

Hello,

I have Gainward Phoenix RTX 3070Ti GPU and just now I noticed these strange artifacts while browsing the internet and on the desktop. After pc restart and driver rollback they are gone for now. A month ago I even had BSOD with this error: VIDEO_SCHEDULER_INTERNAL_ERROR. Any help? Thanks a lot.

System specs:

Ryzen 5600x
Gainward Phoenix RTX 3070Ti 8Gb
MSI Tomahawk B550
Corsair RM 750x
2x8 gigs of 3600 MHz RAM
Win 10 Pro

 

[eidairaman1]

Post the image here directly using attach files button

I am not downloading it

Post detailed specs of your system, saying "650 w psu" is too generic

 

[Daraeon]

Post the image here directly using attach files button

I am not downloading it

Post detailed specs of your system, saying "650 w psu" is too generic

Done

 

[mstenholm]

The only time I have seen this before downclocking the vRAM made it go away but that‘s some AMD generations ago.

 

[Daraeon]

The only time I have seen this before downclocking the vRAM made it go away but that‘s some AMD generations ago.

I rollbacked the drivers from 5.16.xx to 5.12.xx but Im affraid the card could be dying. I have done some undervolting which has been stable for 1+ year, so maybe the 5.16.xx drivers could mess with it? What about the BSOD? Thanks.

 

[eidairaman1]

Try a different port on your monitor/gpu , or try another monitor/cable, otherwise gpu may need rma

 

[siryoink]

I've seen artifacts caused by VRAM going bad, where underclocking would fix it, or from a faulty hdmi cable.

 

[Daraeon]

And is it possible it was caused by the driver itself? Its been 3 days and after rollback everything seems fine.

 

[P4-630]

If you get the same problem again, try if downclocking the GPU memory speed helps, if so, then RMA the GPU because it can't run at it's vram stockspeeds.

 

[lowrider_05]

Looks Like the vram is about to die. How hot is your vram running?

 

[delshay]

And is it possible it was caused by the driver itself? Its been 3 days and after rollback everything seems fine.

Do a little bit of overclocking on the memory, do not increase the voltage . You need to confirm that this was a driver issue.
I have seen these faults in the past & most point to memory fault You may want to monitor memory temperature too.

 

[qubit]

And is it possible it was caused by the driver itself? Its been 3 days and after rollback everything seems fine.

It really should work properly with the latest driver. Dunno why the difference, but it does look like a hardware fault to me. Time for RMA if still under warranty.

You've been running it at stock, or overclocked?

 

[Daraeon]

It really should work properly with the latest driver. Dunno why the difference, but it does look like a hardware fault to me. Time for RMA if still under warranty.

You've been running it at stock, or overclocked?

I have an undervolt @1890mhz core and 0.875V. It was stable for more than one year.

Looks Like the vram is about to die. How hot is your vram running?

This is while FURMARK stress test is running. Its 38 degrees here, normally the temps of undervolted gpu are +- 5 to 8 degrees less.

Btw this is what Forza 5 did with newest driver. No other game did that and since rolling back to 5.12.xx driver Forza is fine. No artifacts.

The GPU is still 2 years in warranty (had 3 years when I bought it and I have it for a year)

 

[P4-630]

Btw this is what Forza 5 did with newest driver.

I don't have that when playing Forza 5 with the latest 516.40 driver.

I'd say your GPU is faulty, there's something wrong with the vram.
RMA.

 

[Daraeon]

I don't have that when playing Forza 5 with the latest 516.40 driver.

I'd say your GPU is faulty, there's something wrong with the vram.
RMA.

How do they replicate the issue, when its not doing it anymore...
Anyway the problematic driver is 516.59

 

[eidairaman1]

I have an undervolt @1890mhz core and 0.875V. It was stable for more than one year.


This is while FURMARK stress test is running. Its 38 degrees here, normally the temps of undervolted gpu are +- 5 to 8 degrees less.

Btw this is what Forza 5 did with newest driver. No other game did that and since rolling back to 5.12.xx driver Forza is fine. No artifacts.

The GPU is still 2 years in warranty (had 3 years when I bought it and I have it for a year)

With that problem, rma the card

 

[qubit]

I have an undervolt @1890mhz core and 0.875V. It was stable for more than one year.

That can also cause a problem. Try the card at stock and see if it works properly with the latest driver.

 

[Zach_01]

I have an undervolt @1890mhz core and 0.875V. It was stable for more than one year.


This is while FURMARK stress test is running. Its 38 degrees here, normally the temps of undervolted gpu are +- 5 to 8 degrees less.

Btw this is what Forza 5 did with newest driver. No other game did that and since rolling back to 5.12.xx driver Forza is fine. No artifacts.

The GPU is still 2 years in warranty (had 3 years when I bought it and I have it for a year)

The card, any card, should be able to run its stock speeds on games without a sweat.
Any other user with your GPU (same brand/model) happen to have issues with same (ver.) drivers?
A new driver release can push the GPU more on load/clocks on some games, so your undervolt is no longer viable. Try to underclock it too if you want better thermals.

Are you running that hairy GPU card killer a lot(often)? Don’t if you do…

 

[Daraeon]

The card, any card, should be able to run its stock speeds on games without a sweat.
Any other user with your GPU (same brand/model) happen to have issues with same (ver.) drivers?
A new driver release can push the GPU more on load/clocks on some games, so your undervolt is no longer viable. Try to underclock it too if you want better thermals.

Are you running that hairy GPU card killer a lot(often)? Don’t if you do…

What you mean by hairy GPU card killer? The Furmark?

 

[phanbuey]

Ram / Ram heatsinks.

 

[Daraeon]

Ram / Ram heatsinks.

care to elaborate? Thanks.

 

[Zach_01]

What you mean by hairy GPU card killer? The Furmark?

Yes that.. Its very abusive on GPU cards.

I believe that @phanbuey was pointing towards improving the VRAM cooling (?)

 

[phanbuey]

So that type of corruption usually appears if either the ram is overheating, or has been baked, in my experience --

Can you run HW info and see what your memory junction is running at?

It's happened to me before on the 2080ti and the 3080 when switching out coolers/ memory pads. On the 3080 it was fixed by applying some k5 pro and im still using that card with no issues. On the 2080ti the chips ended up dying before they were able to be rescued.

@Zach_01 true - I've usually only crashed with those corrupted blocks in games -- also when the old batch of 3090s/3080s with micron chips were biting the bucket they also crashed this way. Although im browsing the net now and the chips are sitting at 72C so if you don't have thermal pads on them they could be baking.

 

[Zach_01]

Though unlikely that VRAM is overheating while browsing the net

 

[Daraeon]

So that type of corruption usually appears if either the ram is overheating, or has been baked, in my experience --
View attachment 255883

Can you run HW info and see what your memory junction is running at?

It's happened to me before on the 2080ti and the 3080 when switching out coolers/ memory pads. On the 3080 it was fixed by applying some k5 pro and im still using that card with no issues. On the 2080ti the chips ended up dying before they were able to be rescued.

@Zach_01 true - I've usually only crashed with those corrupted blocks in games -- also when the old batch of 3090s/3080s with micron chips were biting the bucket they also crashed this way. Although im browsing the net now and the chips are sitting at 72C so if you don't have thermal pads on them they could be baking.
View attachment 255884

This is HW info when idling on desktop. Room temp abou 32-35 celsius atm. Btw I did some corrupted files scan through command prompt and it found some errors. Could that be the cause too?