RTX 3000 Series artifacts, black squares that flicker in desktop and 2D mode

Want to make this post as maybe it will help other users that they think they got a faulty RTX 3000 cards, so if you have this similar issue you can breath relax because your graphic card its ok, the faulty are the nvidia drivers, like any version looks like only afect RTX 3000 series.

There is a issue with artifacts, black/white squares or dots flickering in desktop and 2D mode like more than 1 year old already with RTX 3000 series cards.

At first I tought my RTX 3070 was faulty, so I RMA and I receveid a new RTX 3070 but have same issue. I had the RTX 3070 with the issue from feb 2021 and now from july 2022 I have a new RTX 3080 and after 2 weeks of use, I got the issue again, also in the RTX 3080, so with this card, I tested 3 different cards that all of them has the issue. For me its triggered like 1 per month lets say, and you have to restart PC in order to fix it. I changed DP cable like 3 times aswell just in case.
There is another guy here that posted time ago this issue, he said he RMA his card aswell and the issue triggered for him aswell with the new graphic card.

The issue is very random, there is no chance to reproduce it, the funny fact is that it will have everytime same patron I mean everytime there are some squares that are flickering in desktop, if you move mouse they dissapear, apear again, flickering and most likely the nvidia driver will crash with display has stopped working and recovered. They thend to flicker at same spot every time. Here are some examples of the issue that I found from other users but all of them with RTX 3000 series cards.

1)
2)
3)
4)
5)


I tested bunch of different driver versions, like from beggining of 2021 that I got first time the issue, untill now so there are alot of drivers and all are triggering the issue in certain moment. I did fresh windows install like 10 times , tested windows 10, tested windows 11, nothing works.

 

Underclock the vram and see if it still happens.

 

Does it effects only 30 series? Looks like vram issue. I've never seen anything similar with my card.

 

I only seen that happen when Vram on gpu was dying i remeber my 9800gt did this before it out right died,

I find hard to belive it driver issue effect the 3xxx series and it wasnt not fixed cause there would be alot people complain about

 

Could possibly be your motherboard or even system RAM as well. Try to run a diagnostic like Karhu or something overnight

 

I've had my RTX 3080 since November of 2020 and only encountered this issue once back in November of 2021. When I posted a thread about it here I was told that it was a driver/windows issue. I've never seen this on my RTX 2070S.

What happened was several black squares showed up in the center of my screen while browsing YouTube. After closing the browser they went away. A few minutes later I noticed white squares at the edges of the screen flickering whenever I would scroll through folders.

Currently on Windows 10 with the Nvidia 516.59 driver.

 

I've had major VRAM heating issues with the 3080 Ti Founders Edition (up to 102°C). No artifacts from it yet, thankfully.

The easiest solution is to underclock the memory, which won't hurt performance much. If that doesn't help or you don't want to go that route, you could replace the thermal pads. I've seen great results online for the 3080 Ti FE from others doing this.

I replaced the stock thermal pads on the GPU side with 1.5mm Gelid Extreme pads then cleaned and reapplied thermal paste to the GPU. GPU temps are about the same (around 84-85°C full load, but can weirdly OC more), but the memory temps went down to 96°C or so from 102°C. I've seen much better results online with this mod, so I may need to go up to 2mm pads. If you're comfortable with taking apart the card, the thermal pad mod could be helpful. Do your research and find out what thickness pads you need for your 3070.

 

This looks like @venturi 's 3090 issue,

see if theres any shared similarities between the 2 of your systems.

 

google gpu artifacts and see what's affected

 

What monitor(s) are connected to your graphics card? If more than one, what monitor shows these artifacts? Since this seems to be very rare, we will need logs to start off. Please configure your PC to allow you to force a manual crash as described in the FAQ below:

https://nvidia.custhelp.com/app/answers/detail/a_id/5149

Once configured, leave it like this until you are able to reproduce the squares corruption again. Then use the hotkey to force a manual crash. Once you are back in Windows, upload the crash dump file to a file sharing service and then email me a link to download the file to driverfeedback@nvidia.com.

 

Thanks for replay, the thing is Im not able to reproduce the issue, there is no way I can do it, and as is very rare, some times takes months to reproduce again, I dont know how we will manage this.

Im using single monitor connected with DP cable , tried 3 different already.
Monitor is AOC 27G2U/BK

 

Why is this seemingly acceptable to be buying high-end GPUs to have to manually re-pad for notably better temps? I can understand some intern tossing paste on some $100 Chromebooks haphazardly, but having a single VRAM chip fail due to the vendor not putting in some kind of best-effort to cool a $2000 GPU is stupid, and even more-so that some will argue about warranty if you try to fix it yourself.

Is there any vendor NV or AMD that offers a modern GPU that doesn't need repasted/padded out-the-box for better temps?

 

Agreed. It's ridiculous. NVIDIA really dropped the ball on the 3000 series FE's. This is what I was greeted to when I took it apart.

 

is that paste or pads? it kind hard to tell look kinda like paste to me but that would me insane ammount paste to do that and if there pads? why the hell are they so big, normal pad are size of chip

Quality of products have sure gone to crappers

 

These are NVIDIA's crappy thermal pads. I don't know why they don't use something like Gelid Extreme pads rather than these POS pasty pads.

 

wow i took my 1070ti apparnt to clean and repaste while ago the thermal pads on my 1070ti look better then that, nor did paste on actual gpu look as bad as both look on that. Quality sure has went down. and consider the price that would be insulting

 

the pad quality themselves have never been the problem, the mating pressure has.

 

Did it look like this? I had all your symptoms. I had tickets open, RMAs, had cards replaced with new ones, multiple monitors, cables, etc.

Since the introduction of the 490 series DCH drivers I have seen, across multiple forums, 28 SAVVY users that had the same issue.

For a while the only work around was reverting to the 472 standard driver (made the issues go away).

Also, screenshots can capture it, so its not monitor or cards.




I never got a solution for it from nvidia. No ticket was ever answered with anything usable.
It was never temperature related (GPUs run cool and under heavy load bever get above 69C, the vram never get warmer than 69C). CPU temps are normally 22-24Cm GPU temps are normally 33-41C




I first experienced the polygon issue on the move to dch from 472 to 496 (a forced moved); reverting to the standard series driver such as 472 always solved the issue.


I found a work around for this "soft issue",

--soft because it would come and go on a fresh driver install and after a few reboots it didn't return:



So this is how I got rid of the issue:



For my system (server 2022 data center as workstation)

Extract and copy the Control Panel Client into the nvidia programs directory instead of using the ms store app control panel client.
disable appx (I don't use MS store for anything)
disable Display policy service
set container (nv service) to manual, set recovery to "take no action"
Power off, restart, issue gone

Unique to my setup: I only use control panel during setup of new driver, after that I have the nvidia service set to manual instead of automatic. And if I need to do anything I use inspector. I made an NV control panel shortcut that launches the service if I open nv control panel

I also install only bare driver, no GFX etc.

So in short, no appx, no display policy service, and not running container in the background has kept the problem away.

As a "soft issue" I've come to realize it has to do with the MS/store and nvidia DCH driver and certain hardware combinations. Issue does not manifest in Ubuntu



I don't know if we have any similarities between our two builds, but here is mine for overlap:

2x 3090 Ti RTX Founders Edition & SLI / NvLink bridge
2x 8280L (56/112 cores), Asus c621 Sage Dual socket motherboard, version 2.2
1.5 TB ram. 2933 DDR4 ECC LRDIMMs 1600W digital power supply
(Data drive) 4x VROC Raid 0 Micron 9300 Max (12.8TB each / 51.2 TB volume) and (OS Drive) Sabrent Rocket 4 Plus (8TB). Noctua CPU coolers and A25 fans (black)
Asus PA32UCG-K monitor, using HDMI 2.1 or DP 1.4 DSC, TT SFF case, MS Data Center 2022 & Ubuntu


Problem Solved for me, I hope it helps you

 

I have have personally seen the issue a couple of times on different 30 series cards. Rebooting made it go away and it never happened again on either card.

 

@venturi

thanks for share your experience, I knew it was a driver issue since you are right it startign happaned after 472.12 WHQL but since this is a 2021 driver version and we need to upgrate in order to get good performance in new games we are forced to use latest drivers.

I change also 3 graphic cards and all of them triggered this issue so obiulsy it cannot be the GPU VRAM as people saying.