Games and desktop (GNOME) freezing with NVIDIA

[Aikohh]

Describe the bug

Playing games (usually in fullscreen) causes it to completely freeze seemingly at random. After the game freezes, the entire desktop environment (GNOME) freezes shortly after.
Sometimes after the freezes I get kicked out to the display/login manager, whereas other times I need to power reset the PC to make the PC usable again.
Switching tty with ctrl+alt+F# also does not always work/respond when frozen.

Inspecting dmesg after a freeze occurs, I see some errors/bugs in the output like the following:

[ 2567.811867] CPU: 11 PID: 7294 Comm: gnome-shell Tainted: P    B   W  O       6.9.8-204.fsync.fc40.x86_64 #1
[ 2567.811869] Hardware name: LENOVO 82Y4/LNVNB161216, BIOS M1CN34WW 08/07/2023
[ 2567.811870] ==================================================================
[ 2568.441308] ==================================================================
[ 2568.441311] BUG: KFENCE: use-after-free read in nv_dma_release_sgt+0x29/0x70 [nvidia]

[ 2568.447097] CPU: 10 PID: 7294 Comm: gnome-shell Tainted: P    B   W  O       6.9.8-204.fsync.fc40.x86_64 #1
[ 2568.447099] Hardware name: LENOVO 82Y4/LNVNB161216, BIOS M1CN34WW 08/07/2023
[ 2568.447101] ==================================================================
[ 2580.330429] rfkill: input handler enabled
[ 2581.771320] rfkill: input handler disabled
[ 2591.494070] [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
[ 2594.630080] [drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
[ 2594.678250] rfkill: input handler enabled
[ 2595.806431] rfkill: input handler disabled
[ 2597.453861] [drm:__nv_drm_gem_nvkms_memory_prime_get_sg_table [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Cannot create sg_table for NvKmsKapiMemory 0x000000009682e614
[ 2597.565212] [drm:__nv_drm_gem_nvkms_memory_prime_get_sg_table [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Cannot create sg_table for NvKmsKapiMemory 0x00000000ddcce91a
[ 2597.670667] [drm:__nv_drm_gem_nvkms_memory_prime_get_sg_table [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Cannot create sg_table for NvKmsKapiMemory 0x0000000005b51b71
[ 2599.135940] gldriverquery[30335]: segfault at 200000020 ip 00007f30ead8ce94 sp 00007ffc76b3ee10 error 4 in libLLVM.so.18.1[7f30eaa0c000+3afa000] likely on CPU 5 (core 2, socket 0)
[ 2599.135954] Code: 00 49 89 47 08 48 85 c0 74 41 4c 8b 60 30 4c 8b 68 38 4d 39 ec 74 34 4c 8d 7c 24 10 66 2e 0f 1f 84 00 00 00 00 00 49 8b 04 24 <48> 8b 40 20 48 89 44 24 10 4c 89 f7 4c 89 fe e8 38 8c 00 00 48 89

A full dmesg log is linked in the "Extra information or context" section.

What did you expect to happen?

I expect the games to run normally without freezing up the PC.

Output of rpm-ostree status

State: idle
Deployments:
  ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia:stable
                   Digest: sha256:42149fed522d13a45010df9b562326cdfcd97a019a8f7865953ce692eee5c390
                  Version: 40.20240717.0 (2024-07-17T15:37:37Z)
                     Diff: 21 upgraded, 16 removed, 17 added

● ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia:stable
                   Digest: sha256:30130d99d1d9155ad986d57a93f2d17957ea1948ece71aa03767565b8b13016a
                  Version: 40.20240716.0 (2024-07-16T04:57:23Z)

  ostree-image-signed:docker://ghcr.io/ublue-os/bazzite-gnome-nvidia:stable
                   Digest: sha256:30130d99d1d9155ad986d57a93f2d17957ea1948ece71aa03767565b8b13016a
                  Version: 40.20240716.0 (2024-07-16T04:57:23Z)

Hardware

Extra information or context

I managed to get dmesg output after one of the freezes. It appears the issue is related to NVIDIA drivers.
dmesg-after-freeze.log

In my rpm-ostree kargs I have nvidia-drm.modeset=1 and also tried setting nvidia-drm.fbdev to both 0 and 1 but I still encountered these freezes with fbdev on and off.