Suspend issues despite enabling PreserveVideoMemoryAllocations

[GA251]

NVIDIA Open GPU Kernel Modules Version

590.48.01

Please confirm this issue does not happen with the proprietary driver (of the same version). This issue tracker is only for bugs specific to the open kernel driver.

• I confirm that this does not happen with the proprietary driver package.

Operating System and Version

Fedora Linux 43 (Silverblue)

Kernel Release

Linux 6.18.13-200.fc43.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Feb 19 19:54:01 UTC 2026 x86_64 GNU/Linux

Please confirm you are running a stable release kernel (e.g. not a -rc). We do not accept bug reports for unreleased kernels.

• I am running on a stable kernel release.

Hardware: GPU

GPU 0: NVIDIA GeForce RTX 3060 (UUID: GPU-eed0a0f5-844a-1194-8183-b9e9234e4c1e) - 12GB Variant of a 3060

Describe the bug

Despite PreserveVideoMemoryAllocations being enabled both in modprobe.d and kernel args (seperately), these quirks occur:

1. Suspend takes about 10-15 seconds as opposed to it being about 5 seconds with nouveau,
2. After suspend, after waiting about a minute, there is a high chance that after resume there is corrupt graphics everywhere.
3. Along with that, if the Variable Refresh Rate is enabled, after suspend, every time an app that engages VRR is launched in fullscreen mode, the display glitches out into a seizure mode of sorts until the app is out of full screen mode (where VRR is not enabled in GNOME) or the app is closed.

You can see that the app icons and the wallpaper are glitched. You can also see that apps that are already running are glitched too. But they tend to fix themselves after you navigate to a different page or close and re open the app.

The issue occurs whether or not the Temporary Path is "/tmp" or "/var/tmp", so it's not about the temp path being a tmpfs, though this might be with this GPU being a high VRAM varient?

To Reproduce

1. Install NVIDIA Drivers to Fedora Workstation or Silverblue
2. Add modeset and PreserveVideoMemoryAllocations to the kernel args or to modprobe.d
3. Launch an app that uses the GPU like a game or a web broswer (I have seen this happening most after using firefox, which uses libva-nvidia-driver, which is a compatibility layer that uses CUDA to decode video)
4. Suspend the system while the app is running and in-use
5. Wait about 2 minutes
6. Resume the system

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

More Info

The frequency of this issue increases after an app that uses the GPU extensively is launched and sometimes the issue doesn't occur on the first suspend.

There is also another quirk where after suspend and logging in the display blacks out for about 2 seconds, then turns back on.

I would also like to point out once again that this is a 12GB VRAM variant of a 3060.

[GA251]

Screencast.From.2026-02-28.22-25-20.webm

Another issue that I'm having is that after the failed suspend like this, the suspend function just stops working, pressing suspend in the gnome menu does nothing and pressing the power button (which usually does suspend), falls back to the "Power off" option.

It takes about 2 minutes after the resume for the suspend option to start working again, however there is still about a 1.5 minute latency between pressing the suspend button and the system actually suspending