Youtube, Discord, Visual Studio Code are flickering with black checkered squares/vertical lines. Partial solutions to this issue.

[trparky]

max performance mode

What's that?

 

[mark152]

What's that?

Check my post at page 13 (#303). It's a workaround for the black checkerboard issue. It works both for chrome and edge. If you are only interested in chrome it may be enough to change the angle settings like explained in the previous page (i personally use open gl but there are other options too).

 

[trparky]

Check my post at page 13 (#303). It's a workaround for the black checkerboard issue. It works both for chrome and edge. If you are only interested in chrome it may be enough to change the angle settings like explained in the previous page (i personally use open gl but there are other options too).

Here's to hoping that it'll solve this issue...

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

Hey, Found a game where this happens constantly so it should be fairly easy for anyone owning the game to reproduce.. https://corsixth.com/ (Tested 0.67v) ! You must choose a smaller resolution than what your monitor's native and run in fullscreen.. Depending on the resolution you'll get...

www.techpowerup.com

 

[OneMoar]

open your curve optimiser and do something like this ( attached below)
this solves the problem without needing to enable max performace mode the issue is not enough voltage to sustain the clocks in microbursts this is caused by something in chrome thats causing very short bursts in gpu load which is faster then the nvidia driver will ramp in voltage resulting in minior instablity

a curve like this also results in the gpu staying in boost clocks longer simply because load
you want the high-side of the curve to operate like a clock-step and less like a curve so you need to know what the maxium voltage the gpu will allow is and then raise that to 20-30mzh Above your desired load clocks ( to account for thermals reducing the clock target) in my case thats 1.1v and 2040Mz at >=70c give or take a few degrees

this is a regression not a new bug both the 1000 and 2000 series cards had this exact same issue something is janky with the way chrome loads up and nvidias powermisering at low gpu loads
Edit: fixed screenshot

for those that don't understand how co works this effectively adjusts the range that the powermiser can operate in without giving up performance or power consumption where it matters

 

[mkdr]

open your curve optimiser and do something like this

This is nonsense. This would indicate it is 100% Nvidias fault and they could fix it with a driver. The fix though is inside the Windows OS / DWM mostly.

Also this would totally ruin undervolt, which obviously everyone wants to use.

 

[OneMoar]

This is nonsense. This would indicate it is 100% Nvidias fault and they could fix it with a driver. The fix though is inside the Windows OS / DWM mostly.

Also this would totally ruin undervolt, which obviously everyone wants to use.

that is a undervolt its -350mhz at the left end of the curve -60 at the middle and +100 at the peak voltage range (which is a overclock if you don't want to overclock you can simply leave the top end of the curve at +0)

there is no static unvolting anymore the only controls you have are the clock range and temp gone are the days of setting a static frequency all you can do is 'suggest' to the driver what you want the card todo
curves are not calculatated from a fixed VF-curve like the name would implie at idle all you are adjusting is the effective peak range(boost clock for a given point in the vf-thermal-load calculation ) the gpu will still idle down to ~200Mhz Max perf mode just tells the gpu to goto its maxium clock for the gpus desired point in the curve

its a work around untill whatever hardware acceleration shenanigans are resolved by _ALL_ involved parties

also the assumption that everybody wants to undervolt is a blanket and incorrect statement

 

[wanderingpanda]

So what does that mean? Because it's more restricted shouldn't that mean it's in the obviously capable hands of Nvidia? But it also sounds like because of that the fault is more likely on them.

Then again that is what's dictated, and windows fails to cooperate. Does this happen on other platforms?

 

[Sora]

This issue also happens at max clocks, it just happens less because the timing required to provoke the race condition in the compositor is minimized by processing the compositor faster.

 

[xorbe]

This smells like the GTX 660 issue from years gone by ... they would have it fixed in one driver, and "break it" again the next driver ... pretty sure it was a hardware issue where the work-around needed to be fine-tuned for each driver build.

 

[Wezelkrozum]

Just a heads up for the people that were using the workaround in my previous post #post-5104215

 2. And/Or add --disable-direct-composition-video-overlays to a browser shortcut

According to this reaction on my Chromium bugreport that flag has recently been changed in Chromium based browsers to:

 --disable_direct_composition_video_overlays=1 --disable_direct_composition_sw_video_overlays=1

 

[rithra]

First of all, hello,

I am having a black square problem, I am currently using Asus Dual GeForce RTX 3060, I was using 1050 TI before, I switched to 3060 at the beginning of this year at the end of January, I was already having problems while using 1050 TI, I bought a new card, the problem still persists that there is a problem with the video card. I don't think so because I tried windows 10 and 11, I formatted and both have the problem, but in normal time it was only in chrome, in the last 2 weeks it started to happen in discord, will this situation start to spread or will this problem be solved by nvidia, this situation bothers me.
I would be grateful if you could give me some information

 

[Sora]

First of all, hello,

I am having a black square problem, I am currently using Asus Dual GeForce RTX 3060, I was using 1050 TI before, I switched to 3060 at the beginning of this year at the end of January, I was already having problems while using 1050 TI, I bought a new card, the problem still persists that there is a problem with the video card. I don't think so because I tried windows 10 and 11, I formatted and both have the problem, but in normal time it was only in chrome, in the last 2 weeks it started to happen in discord, will this situation start to spread or will this problem be solved by nvidia, this situation bothers me.
I would be grateful if you could give me some information

Don't cross post, and read the thread.

 

[rithra]

Don't cross post, and read the thread.

I read all the topics, there are places I do not understand, but I could not find a solution.

 

[saulob]

There's no solution yet, it's a Windows bug, we are waiting for the Microsoft update for it

 

[flynsquirrel]

Has anyone noticed this in Steam? Been troubleshooting an issue with my GPU and after sucessfully getting my computer up and running I'm now seeing this where I wasn't before.

 

[mkdr]

Has anyone noticed this in Steam

yes same issue, steam uses chromium

 

[Sora]

kb5035942 possibly fixes this.

 

[Wezelkrozum]

kb5035942 possibly fixes this.

Sadly no. I installed it 2 days ago, but still experience 100+ artifacts daily.

 

[mkdr]

kb5035942 possibly fixes this.

I have tested it for the past 1 hour and it seems to fix the issue. I dont get jumping elements and black bars anymore when I am on twitter.com and move mouse cursor over elements in chrome while a video is playing. I could reproduce the issue before with 100% chance by just go on twitter, have a twitch.tv video playing in another tab, and then move the mouse cursor over addon icons in the chrome bar on the top. Every time the mouse cursor enters a little animation with mouseover like making a chrome addon icon background color gray, it would produce jumping and flickering elements on the twitter site or black cardboards.

 

[PunchyXTuna]

Sadly no. I installed it 2 days ago, but still experience 100+ artifacts daily.

KB5035942 (os build 22621.3374 or 22631.3374) was not the one that worked for me ; Its the 24H2 or 26090.1 os build That did , If you sign up to window's insider program and pick the Dev channel and go check for updates it should automatically install, Im pretty sure you can also find the iso here as long as you are signed up to the insider program : https://www.microsoft.com/en-us/software-download/windowsinsiderpreviewiso just make sure that the os build that starts with 26xxx.

 

[saulob]

kb5035942 possibly fixes this.

Why? Do you have more info on that? I was unable to find it

I mean, any list of fixes showing it, any change log

Sadly no. I installed it 2 days ago, but still experience 100+ artifacts daily.

Me too. Same problem. Still happening

Windows 11, 22H2, build 22621.3374

 

[mkdr]

I still stand by what I said. Issue seems to be fixed since kb5035942 on Windows 11 23h2 22631.3374 on my PC in Chrome 123.0.6312.86 with Nvidia drivers 551.86. Cant reproduce the issue anymore since 2 days now. I dont use Discord though or Visual Studio, just speaking about Chrome.

 

[Sora]

I still stand by what I said. Issue seems to be fixed since kb5035942 on Windows 11 23h2 22631.3374 on my PC in Chrome 123.0.6312.86 with Nvidia drivers 551.86. Cant reproduce the issue anymore since 2 days now. I dont use Discord though or Visual Studio, just speaking about Chrome.

the kb update might be the partial fix that reduced but did not 100% eradicate the issue in 22635, there might be other factors that are fixed in canary such as multi-display, with mpo's enabled.

There is a chrome (the gui, not the browser) checkerboarding issue in the CEF engine that is unrelated to the dcomp issue where it occurs on updates to the scrollbars, so there might be still a chrome issue afoot to be fixed.

 

[mkdr]

@Sora can you link the chrome bug for that. I also dont use multi display btw, I think I had enabled MPOs though some time ago again.

 

[Sora]

The scrollbar issue is https://issues.chromium.org/issues/40680264