Yes, the workaround with the kernel parameter doesn’t work anymore with 535.43.02 driver.
I’ve tried the following :
Downgrade to the 525 driver. With and without nvidia_drm.modeset=1, i’ve no flickering at all.
Upgrade to the 530 driver. Without nvidia_drm.modeset=1, the flickering occur. With the parameter, no flickering.
Upgrade to the 535.43 beta driver. With and without the kernel parameter, flickering occur.
Yes, it appears when the p-states fluctuates and more frequently between the two first p-states (P0 and P1)
If it really bothers you, the only way to reduce flickering is to prevent the GPU from switching too much between p-states. Personally, I’ve lowered the refresh rate of my screen to 100hz instead of 144hz. I’ve also disabled my compositor (Picom). With this, the flickering hardly appears anymore.
Fortunately, in game, I have no flickering. This seems normal because the GPU constantly uses the p4 state, which corresponds to the “maximum performance” mode.
However, I advise you not to use the “maximum performance” mode every time in the NVIDIA control center because your GPU has its video memory constantly at its maximum frequency and it consumes + heats up a lot.
40w vs 12w idle.
Finally, I reread the previous topics dating from April and I suspect that only GPU based on the Turing architecture are affected by this problem.
Today I have used my pc with the max performance mode the whole day and the flickering seems gone.
To me this seems the best option because I like to use 144 hertz and the compositor (I also use picom).
Sure the higher power consumption is bad but I hope that NVIDIA will fix this soon™…
Or if it reappears and gets unbearable I will just reinstall 525.116.04 I guess.
PS:
To anyone reading, you can set the power state with: nvidia-settings -a "[gpu:0]/GPUPowerMizerMode=1"
and check if it is setted correctly with: nvidia-settings -q GpuPowerMizerMode
Or you can use nvidia-settings gui application, but don’t use them both since they can interfere with each other (e.g. you set from terminal the performance mode, then you open the gui app and it reverts the setting to auto).
If I activate the maximum performance mode, my GPU (an RTX 2080) consumes 40 watts. As a result, its temp gradually rises to 60 degrees and then the fan activates until the temperature drops to 47 degrees and so on.
That’s why I prefer to put the screen at 100 hz. I can stick with adaptive mode and the card only draws 12-13 watts. Moreover, the clock frequencies are at their minimum.
It is likely that your GPU is better cooled than mine, so if you don’t have excessive heat, you can stay in this mode.
Reinstalling 525.116.04 is clearly the best solution until the problem is fixed. However, the kernel will evolve and new versions will be released. There will come a time when 525s will no longer build.
Additionally, 535 fixes a freezing issue with AC Valhalla that occurs when GPU usage hits 90%. Going back is not an option for me unfortunately.
I also hope that this issue will be resolved soon. But knowing that it was first declared with the release of the 530 and if it is true that it really only concerns cards with the Turing architecture, I fear that it will take longer to be corrected.
Fingers crossed.
Thank you for the command lines to set PowerMizerMode. I was using nvidia-settings GUI before you mention it.
By default I also have the fans activating over 60 degrees, when it happens they start full speed so it isn’t pleasant.
If it was only when playing a game would have been bearable, but it does it even using the browser (even with adaptive mode, surely it is worse as you say with performance mode).
So sometime ago I have decided to use nfancurve to set my fan curve: I have a minimum fan speed of 25% (under 60 degrees), which means ~850 rpm and 45 degrees celsius (pretty much constant using desktop).
In any case let’s hope they fix this, because it is not good at all!
I have installed the new “Production Branch Version” 535.54.03 driver, it has this issue.
How can a production branch version have a glaring issue like this?
Right now every downloadable 5xx driver is affected by this issue, not bad.
This issue has existed since version 530 was released.
I reread the threads on this forum talking about the flickering problem. I also reread those from other forums (manjaro, endeavourOS). Absolutely everyone having this problem has a card whose hardware architecture is Turing (GTX 1660, RTX 2060, 2070s, 2080, 2080s and 2080TI).
The first topic was created at the end of March. We are in June and it is still there with the 535 betas. Unfortunately, there was little chance that this would be resolved for the release of the 535 stable.
If a version fixing this problem should be released, I don’t think it will be before the next major release; maybe 540. But we’ll probably have to wait until September for that. Turing is an architecture equipping GPUs released in September 2018. 5 years is a decade in computing. We’ve the RTX 3000s and now the RTX 4000s. Bugs affecting the RTX 2000s may not be the priority.
In the meantime, we have to use workarounds to limit flickering (as mentioned above) or revert to the 525 drivers. However, if gaming is a priority, we’ll have to do without the new vulkan extensions and performance improvements.
Yes, I have read your previous messages and yes, also in my experience it is since driver 530 that we have this issue… but my point of view is pretty different because we are talking about:
Normal video output usage being bothersome without workarounds;
An architecture (Turing) which is currently supported and not a legacy one (the years don’t matter);
We are talking about the production branch (which should be the smoothest in this regard).
Wanted to confirm that I’m suddenly seeing a similar but worse issue – the whole screen flickers completely to black and back each time this occurs. These interruptions last roughly a second.
I am on Arch Linux, which recently updated to 535.54.03. Issue started right after update. I’m using RTX 2070 Super and nvidia-drm.modeset=1. My resolution is 3840x2160 @ 144 Hz.
I can confirm that the PowerMizer workaround listed here does work, but my GPU power draw more than doubles from around 25W to 55W. Not ideal at all. I can reproduce the issue in a matter of minutes when switching back to the adaptive setting.
The power switching (and a few other things) were speculated about with the 530 drivers as everyone frantically tried to figure out wtf was going on but it was all bumkis. If it worked it was either coincidental or just decreased the frequency making it seem fixed until BLAM, there it is.
Fingers crossed they fix it this time and perhaps fix the search on here since this post is a few days before mine but didn’t show when I searched before reposting about the issue returning for the 535’s.
I don’t notice anything when playing games, everything is ok on the top screen as well. But the desktop (KDE NEON 22.04) does not run so smart anymore. There it did not before.
Driver directly from the Ubuntu Repoitory 535.54.03
Nvidia GTX 4090
Hi All,
I have seen flickering of entire screen on my setup locally and we are actively debugging the issue.
Bug 4164132 has been also raised for the same.
Hello guys, i was one of the guys who created a Topic about this Black Flickering few month ago and even with the workaround with nvidia_drm.modeset=1 that could solve the issue for a while… the bug comes back with 535.54.03… atleast its not this annoying as few month ago but it still a pain to see it again :(
I also removed today nvidia_drm.modeset=1 from grub but no changes.
Can we PLEASE FOR THE LOVE OF GOD get a fix for this? I know it only affects Turing GPUs and it’s not a high priority but this is something that was working just fine and now it’s completely broken! This should have been fixed straight away! My distro doesn’t allow me to downgrade to older drivers and I’m at my wits end here.
Luckily with the power states set to performance mode I have this kind of flickers very rarely… I can bear it but I don’t want to imagine using my PC without this setting applied.
I think that to certain people a constant flicker like this one could even cause headaches.
Furthermore as said there were already reports months prior to my post here, pointing to some issue in the drivers.
The response from NVIDIA acknowledging this issue on daily PC usage, starting to debug it just one week ago is really bad from a market leader: they waited for every possible user with a Turing card and an high refresh monitor to have this issue before moving.
This slow response caused the drivers to get pushed as stable on many repositories, without users being able to downgrade: NVIDIA’s site itself doesn’t allow to download 525 drivers anymore! and even when this will be fixed it will take some time to have it as a stable release on repositories.
I hope this gets fixed soon and that it is just a one off: issues like this can’t be managed like it’s happening here.