I’ve been running bazzite-nvidia-open for a while, and with the Fedora 42 builds I started noticing an occasional (every few days) hard freeze that requires a reboot. This has happened even when I’m just watching a video (on Firefox). There does not seem to be any slowing down or stuttering before this happens. This problem has persisted even with the 42.20250603.1 build with the 575 driver. I dual boot to Windows and the freeze does not seem to happen there, which leads me to believe that it is not a hardware issue.
I was wondering if anyone else has had this problem. I attempted to rebase to the last 41 build to see if that would help but there but the deployment didn’t take (perhaps deserving a separate post if this is indeed supported).
The system has an NVIDIA 4070 Ti Super graphics card with two monitors, and an AMD Ryzen 7 7800 X3D processor.
1 Like
I’m using my own image based on kinoite. After a while (like 2 days), nvidia stops working. Didn’t do anything about it as I don’t have time and it’s quicker to just reboot
This started happening with the new 575 drivers, I don’t recall having this issue before
Working fine on my PC, Aurora DX nvidia-open (stable-daily), AMD Ryzen 7 5700X, NVIDIA RTX 3060 and NVIDIA open driver 575.57.08.
I had problems with suspend before switching to Aurora from Ubuntu 24.04 LTS, but I don’t remember having a freeze.
I’m on Bazzite 42 with nvidia-open driver 575.64 and I keep getting these freezes too for which I have to force-restart my system. This happens quite too often and not even while playing games.
Journalctl tells me this:
[drm:nv_drm_gem_alloc_nvkms_memory_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x…] Failed to allocate NVKMS memory for GEM object
I have an RTX 3060 Ti.
I also used Nobara before and I didn’t have this issue. I’m thinking it’s working on Nobara maybe because it’s using the proprietary(?) akmod driver instead of the open one.
Also, Nobara was on the latest 6.15.4 kernel and nvidia(prop) version 575.153 whereas bazzite is on 6.14.6 kernel and nvidia-open 575.64.
Someone suggested me I select the Legacy GPU 10- series card and download that image for which I’d get the nvidia-prop driver. I see that I can also do the same by rebasing to bazzite-nvidia:stable. I might try this and see if it becomes better.
After almost a month I had the possibility to do an update, I’m on holiday and although having an internet connection it is one which is on the clock. Downloading a few GB’s is simply too much, so I went for a coffee with free wifi. That explains the large difference in deployment numbers.
Aurora version is: ostree-image-signed:docker://ghcr.io/ublue-os/aurora-nvidia:stable
The original deployment is Aurora version 42.20250608.1 (2025-06-08T06:31:55Z),
with kernel 6.14.5-300.fc42.x86_64,
and Nvidia driver is 575.57.08.
This one works great, no problem at all.
The one with the problem is Aurora version 42.20250703 (2025-07-03T00:58:41Z),
with kernel 6.14.9-300.fc42.x86_64
and Nvidia driver is 575.64
Here regularly the system freezes no matter how long I wait for it to become responsive again, so I use the Power Off button and start over.