Summary
Intermittent display initialization failure on KDE Plasma Wayland with NVIDIA.
Two Dell DisplayPort monitors (DP-1, DP-2) sometimes come up in fallback low resolution (640x480, occasionally appears as 540x480) after either:
- normal boot/login, or
- suspend/resume.
They remain unusable until physically disconnecting/reconnecting the affected monitor cables.
Environment
- OS: Fedora release 44 (Forty Four)
- Kernel: 7.0.4-200.fc44.x86_64
- GPU: NVIDIA GeForce RTX 5070 Ti (PCI 0000:2d:00.0)
- NVIDIA driver: 595.71.05
- Desktop session: KDE Plasma Wayland (kwin_wayland)
- Plasma version: plasmashell 6.6.4
- KWin version: kwin 6.6.4
- Package versions:
- plasma-workspace-6.6.4-1.fc44.x86_64
- kwin-6.6.4-2.fc44.x86_64
Display setup
- DP-1: Dell U2312HM
- DP-2: Dell U2312HM
- DP-3: MSI MPG 275CQRXF
- HDMI-A-1: Samsung panel
- No KVM or adapters on affected Dell monitors.
Expected behavior
DP-1 and DP-2 should consistently initialize with real EDID/native modes after boot and after resume.
Actual behavior
DP-1 and DP-2 are detected as connected but expose only fallback low resolution; desktop layout becomes unusable.
Software re-probing does not recover; physical replug does.
Observed failure modes
A) Resume failure
- Around resume, logs include:
- “Atomic modeset test failed! Permission denied”
- “EDID colorimetry … is invalid”
B) Normal boot failure
- On some normal boots/login, the same fallback behavior appears without suspend.
- Same EDID/modeset warning pattern appears in logs.
Evidence collected
Broken-state bundle:
- nvidia-resume-edid-bug-2026-05-15-084917.tar.gz
- Includes nvidia-bug-report.log.gz and full journal snapshots.
- Broken EDID decode for DP-1/DP-2:
- Manufacturer: NVD
- Model: 0
- Made in: 1990
- Fallback timing (640x480 only)
- Broken kscreen state shows DP-1/DP-2 fallback-only mode.
Working-state bundle (after replug):
- nvidia-resume-edid-working-2026-05-15-085953.tar.gz
- Working EDID decode for DP-1/DP-2:
- Manufacturer: DEL
- Model: 16499
- Correct monitor identity/modes restored.
Additional note
A normal cold boot often works, and physical replug consistently fixes the issue mid-session.
This strongly suggests intermittent EDID/modeset/link-training handling in the driver stack rather than static compositor layout/config corruption.
Mitigations attempted (no in-session recovery)
- kscreen-doctor disable/enable DP-1/DP-2
- DPMS off/on
- udevadm trigger --subsystem-match=drm --action=change
- restart nvidia-resume.service
Workaround
Physical disconnect/reconnect of affected DP monitor(s).
Attachments
I recently had EXTREMELY similar behavior with a monitor locked at 640x480 and undetectable in Linux (Bazzite, based on Fedora). I thought my monitor was failing. Bizarrely, even a reboot did not fix it, but once I started swapping cables it worked again.
However, this was with driver 595.45.04, so I wonder if the issue is even older than 595.71.05.
Same bug — RTX 5070 Ti, Bazzite (Fedora 44), KDE Wayland — triggered by monitor input-source switching (intermittent)
I’m hitting what appears to be the exact same bug on a very similar setup.
Environment
- OS: Bazzite (atomic Fedora 44 base), Wayland-only image
- GPU: NVIDIA GeForce RTX 5070 Ti
- Driver: nvidia-open 610.43.02
- Desktop: KDE Plasma 6.6.5 / KWin 6.6.5 (Wayland session)
- nvidia-persistenced: active / enabled
- nvidia_drm:
modeset=1, fbdev=1 (both confirmed active)
Display setup
- 2× identical Iiyama PL2770Q (2560×1440 @ 180 Hz), connected via DisplayPort (DP-4 and DP-5)
- Scale 1, no VRR, no HDR
- Both monitors are shared with a second PC via the monitor’s built-in input-source switching (no KVM, no adapters)
Trigger
The failure occurs when I switch a monitor’s input source to my second PC and then switch back to the Bazzite machine. On return, the display layout breaks: the affected DisplayPort output drops to a fallback mode and the desktop arrangement is lost.
Symptom / evidence
When the bug triggers, the affected output’s EDID is decoded as a phantom display: NVD 0 0 0 0 0, 640×480 fallback timing. This phantom output gets written into KWin’s kwinoutputconfig.json alongside the real monitor, and persists/regenerates at boot. This matches the broken-state EDID decode described in this thread (Manufacturer: NVD, Model: 0, 640×480 only).
Intermittent behavior
The failure is random / non-deterministic. The same input-source switch sometimes restores the layout correctly and sometimes triggers the fallback — there is no consistent reproduction. A cold boot sometimes comes up clean, sometimes already broken. This intermittency strongly suggests an EDID / modeset / link-training race in the driver stack rather than static compositor config corruption.
Mitigations attempted — none recover in-session
kscreen-doctor output enable/priority/position commands — no effect (also note: kscreen-doctor appears broken under Wayland here)- TTY switch (Ctrl+Alt+F3 → F1) — does not restore
- Deleting
kwinoutputconfig.json + reboot — phantom NVD output regenerates
- nvidia-persistenced +
modeset=1 + fbdev=1 already in place (workaround from a related Bazzite issue) — does not prevent this bug
Working workaround
Physically unplugging and replugging the affected DisplayPort cable, full reboot, consistent with this thread.
This is a significant usability problem for any multi-PC setup that shares monitors via input-source switching. Adding my report to raise priority, since this is now confirmed on RTX 5070 Ti / Fedora 44 / KDE Wayland across multiple users.
I just had this behavior yesterday, also on Bazzite.
In my case I managed to recover without rebooting the system. Instead, I unplugged both the DVI and the power cord from the monitor, waited a few seconds, and plugged both back in. At that point, the monitor was detected as the correct model again.