[DRM,NVIDIA] - Embedded Gamescope: Steam notifications/overlay cause artifacts and flickering

[alneor-ellyuan]

Is there an existing issue for this?

• I have searched the existing issues

Are you using any gamescope patches or a forked version of gamescope?

• The issue occurs on upstream gamescope without any modifications

Current Behavior

After starting gamescope and Steam from tty, when in game, receiving a Steam notification (e.g. an achievement) causes artifacts and corruption to appear on screen. The same issue is also caused by simply displaying the Steam overlay (in game only). It looks like the game and the steam overlay are competing for focus: moving the mouse fixes the issue (at least, until the mouse pointer dissapears from inactivity). The corruption stops if the steam overlay is closed.

If a graphical session is started on a second tty (e.g. sway), the artifacts carry from the gamescope tty to the other. The corruption seem to be made from from the images of both ttys. Moving the mouse on the other tty might fix the display, not always. By alterning ttys back and forth, I can eventually bring nvidia-drm to crash:

[drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
[drm:nv_drm_atomic_commit [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Flip event timeout on head 0
[drm:__nv_drm_semsurf_wait_fence_work_cb [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to register auto-value-update on pre-wait value for sync FD semaphore surface

I tried various gamescope flags (e.g. --immediate-flips, --force-composition), to no avail. And while these logs were captured while playing my latest personal obsession, the issue appears in the other games I tried.
P.S. iirc, gamescope segfaulted after I closed steam with steam -shutdown at the end of my tests, so that seems like another issue completely.

Possible similar problems include:

• Steam Notifications change current window and make games flicker #1909: has flickering caused by notifications with wayland backend (not drm), yet not artifacts (caused by nvidia?);
• [Nvidia] Black screen when HDR option is enabled #1593: has artifacts on drm backend with nvidia, but caused by hdr, not the Steam overlay.

Steps To Reproduce

1. Launch gamescope in tty with gamescope -e -- /usr/bin/steam -tenfoot (actually gamescope --debug-focus --debug-layer -e -- /usr/bin/steam -tenfoot for the logs, but the issue appears without the debug flags).
2. Start a game.
3. Make the Steam overlay appear ("shift+tab", or with controller): corruption appears.
4. (Optional) Switch to another TTY with a graphical session: the corruption is still there.

Hardware information

- Distro: "Arch Linux"
- CPU: AMD Ryzen 7 7800X3D
- GPU: NVIDIA GeForce RTX 4090
- Driver Version: NVIDIA 580.82.7.0

Software information

- Desktop environment: sway 1.11
- Session type: wayland
- Gamescope version: 3.16.16
- Gamescope launch command(s): `gamescope -e -- /usr/bin/steam -tenfoot`

Which gamescope backends have the issue you are reporting?

• Wayland (default for nested gamescope)
• DRM (default for embedded gamescope, i.e. gamescope-session)
• SDL
• OpenVR

Logging, screenshots, or anything else

First run: in game, artifacts appear when the Steam overlay is called; and dissapear when the mouse is moved.
tty1-gamescope.log
tty1-nvidia-bug-report.log.gz
https://github.com/user-attachments/assets/c6d6b3e4-cd24-4635-876e-9e87eece9ef9

Second run: in game, artifacts appear when the Steam overlay is called. By moving to another tty with sway loaded, the artifacts are still there.
tty1-2-gamescope.log
tty1-2-nvidia-bug-report.log.gz
https://github.com/user-attachments/assets/0272381b-cc7d-4b99-a3ce-af32ab115244

[zgibberish]

I'm having the exact same issue!

distro: Arch Linux
kernel: CachyOS kernel
GPU: Nvidia RTX 2060
Drivers: linux-cachyos-nvidia-open 6.16.4-4 (580.82.07)

Only happens when started from TTY, works fine on Niri wayland compositor.

And the flickering issue appears on other apps too, not just Steam.

[zgibberish]

Hey, this might be an NVIDIA issue, downgrading to version 575 of nvidia drivers fixed the flickering completely for me.

(It was hard to find older package versions for kernel, kernel headers, nvidia modules and nvidia-utils tho)

[alneor-ellyuan]

You are right, the issue seems to dissapear after downgrading to 575: thanks for the finding! I made a post on nvidia forums to report the bug, we'll see how it goes: https://forums.developer.nvidia.com/t/580-release-feedback-discussion/341205/427

[TypoMustakes]

You are right, the issue seems to dissapear after downgrading to 575: thanks for the finding! I made a post on nvidia forums to report the bug, we'll see how it goes: https://forums.developer.nvidia.com/t/580-release-feedback-discussion/341205/427

Have you ever receive any help regarding this? I don't want to run a legacy driver for eternity just because of this.

All I see in the discussion is that they replied with "Thank you for reporting issue, do you see this issue only if display modes are above 2560x1440p@120hz with HDR enabled or without HDR and low refresh rate as well?", and then left it at that.

It would be typical from Nvidia support to leave us stranded, but I hope they don't

[TypoMustakes]

FYI: Somebody tried to resurrect the issue in the 595 release discussions, to no avail: https://forums.developer.nvidia.com/t/595-release-feedback-discussion/362561/151

This was in March, half a year after the first report, and was left just as unanswered as that one. I... really dislike Nvidia

[defmetal]

Reproduced on different hardware, with new isolation data that narrows the cause.

System

• GPU: RTX 5090 (GB202), nvidia-open 610.43.03 (DKMS)
• gamescope 3.16.23.2 (git 20260628, Nobara 44 package), DRM backend, gamescope-session-plus
• Kernel 7.1.4, Fedora/Nobara 44
• Display: 5120x2160@165 over DP 2.1, SDR (HDR disabled for these tests)
• Client: Steam stable 1784778118 AND beta 1784934043 — both affected

Symptom: garbage in the background of Steam's bottom-right overlay surfaces
(gamepad button legend / QAM edge / notification region) in Big Picture under
embedded gamescope. Text/glyphs composite correctly ON TOP of the garbage.
Follows into games. Content of the garbage varies per boot (specks, confetti,
tiled teal blocks — the tiled repeating pattern suggests a tiling/stride
misinterpretation rather than random uninitialized reads).

Isolation matrix (each row one variable):

Configuration	Result
Same client, same mode, Big Picture under KWin (desktop Wayland)	clean
Embedded gamescope, defaults	corrupt band bottom edge when idle; clean while cursor active
+ --force-composition	main plane fixed; overlay-surface background still corrupt
+ WLR_NO_HARDWARE_CURSORS=1	no change (cursor plane exonerated)
Steam "GPU accelerated rendering in web views" OFF (CPU CEF)	clean, unusably slow at 5K2K
steam -cef-disable-gpu-compositing (GPU raster, shm handoff)	clean, but overlay blinks ~1 Hz
Steam client beta 1784934043, full GPU path	corrupt (same as stable)
Driver rollback to nvidia-open 595.84 (production), same config	tiled garbage gone; only a small translucent black box remains on the overlay surface (alpha not blended — matches NVIDIA forum 369869)
595.84 WITHOUT --force-composition (direct scan-out)	dead-pixel-style specks return; black box unchanged — scan-out corrupts on both branches, and the black box is independent of the composite path

So: any path where steamwebhelper's surface reaches gamescope without GPU
buffer sharing is clean; the GPU-shared path is corrupt only under gamescope,
not under KWin. Rolling the driver back to 595.84 (production branch)
with an otherwise identical config eliminates the tiled garbage entirely —
610.43.03 is a regression. On 595 a milder residue remains: the overlay
surface composites as a small translucent black box (alpha ignored), i.e.
NVIDIA forum 369869's symptom exactly — pixel data imports correctly on 595,
alpha handling is still wrong on both branches. Note this is likely the gamescope-side face of a longer-running
NVIDIA+CEF dmabuf issue: steam-for-linux #10537 (Feb 2024, driver 550, desktop
Wayland, no gamescope) shows the same timing-dependent web-view corruption,
where slower menu renders corrupt more often and a window resize (buffer
reallocation) temporarily clears it — matching the per-boot variability seen
here at 165 Hz. Consistent with the OP's report that --force-composition
alone does not fix it. Logs also show xwm: got the same buffer committed twice, ignoring (#1383) and zero-modifier failures for AB48/XB48 on this
driver, in case those are related to modifier negotiation.

Happy to test patches or provide drm.debug logs.

[SimpleHeuristics]

Replying to #1964 (comment)

So the deck version (Gamescope session) of Bazzite 44 is also releasing soon and on the 610 drivers it is showing the artifacts on the bottom bar of the steam UI at 4K 240HZ and 120Hz in SDR or HDR with the UI.

However if HDR is enabled and I launch a game that has native HDR support such as Cyberpunk or Assassin's Creed Shadows it cleans up the artifacting at the bottom of the screen unless I bring up the overlay use the performance monitoring overlay.

For non HDR native games where gamescope will do its inverse tone mapping this unfortunately doesnt change anything and the visual artifacts remain on the bottom of the screen.

1440p with HDR is fine with the 610 drivers.