Component: Linux / Graphics / NVIDIA proprietary driver
Distribution: EndeavourOS (rolling)
Kernel: 6.17.7-zen1-1-zen
Hardware: Lenovo Legion 7 16ACHg6 — AMD iGPU + NVIDIA GeForce RTX 3080 Mobile
Driver versions:
- Last working version: 580.82.09-1
- Regression present starting version: 580.95.05-1
- Latest version also failing: 580.105.08-1
Bug description:
When connecting an LG TV (with HDMI input supporting 120 Hz) to the laptop’s HDMI output (managed by NVIDIA dGPU) under hybrid GPU configuration (AMD iGPU + NVIDIA dGPU), the graphical environment freezes/crashes (X server terminates) and only TTY remains accessible. The internal panel eDP drops out.
This behaviour begins with driver version 580.95.05-1 and persists in 580.105.08-1, but did not occur with version 580.82.09-1.
Steps to reproduce:
- Boot system with driver version 580.95.05-1 (or later).
- Connect LG television via HDMI at 120 Hz.
- Observe freeze or crash of graphical session.
- Revert to 580.82.09-1 → everything works.
Expected behaviour:
With 580.82.09-1: external 120 Hz display works, internal panel remains active, no crash.
Actual behaviour:
With 580.95.05-1 and 580.105.08-1: connecting external 120 Hz display causes session freeze/crash.
Logs/Diagnostics attached.
Request:
Please investigate regression after 580.82.09-1 affecting hybrid GPU systems + HDMI 120 Hz displays. A patch/bug-fix would be appreciated.
Attachments:
Thanks for the link.
Yes, this does appear to be the same issue as bug 5622970. The symptoms and affected hardware setup (Lenovo Legion 7 + RTX 3080, HDMI output at 120 Hz, hybrid GPU configuration) match exactly.
For reference: the problem is still present in the latest driver version 580.119.02-1 as well. The regression persists beyond 580.95.05-1, and the issue has not yet been resolved.
Reverting to 580.82.09-1 remains the only known workaround on my system.
Hopefully this helps with correlating the reports and prioritizing a fix.
Update (still broken on latest drivers) + request for assistance
I’m posting an update because the regression is still present and I’m increasingly frustrated by the lack of progress or concrete guidance.
In this same thread I already provided full diagnostics (nvidia-bug-report.log.gz, journalctl snippet, Xorg log snippet, README).
Since then I have continued testing newer driver stacks. Most recently I tested the nvidia-open driver line (590.48.01) and the result is the same: connecting my LG TV via HDMI triggers a crash.
Regression Summary (confirmed)
System / Setup
Laptop: Lenovo Legion 7 16ACHg6
GPU: NVIDIA RTX 3080 Mobile (hybrid setup with AMD iGPU)
Distro: EndeavourOS (rolling)
External display: LG TV with HDMI input supporting 120 Hz
Last known good:
- 580.82.09-1
First known bad:
- Regression starts at 580.95.05-1
Still failing:
- 580.105.08-1
- 580.119.02-1
- 590.48.01 (nvidia-open)
Trigger:
Plugging in the LG TV via HDMI (hot-plug) in hybrid GPU configuration.
Observed result:
Graphical session crash/freeze and loss of usable display state (as described in the original report).
With the newest stack I’m testing, the failure is still severe enough that it effectively makes HDMI output unusable on my system.
Workaround:
Reverting to 580.82.09-1 remains the only reliable workaround.
Request: can anyone please help with a fix / patch / debug instructions?
At this point I’d like to ask:
- Can an NVIDIA developer confirm whether this report is indeed tracked under 5622970 (or another bug ID)?
- Is there any patch, test build, or specific debug procedure you want me to run to produce more useful traces?
If additional logging is needed, please provide exact kernel/module parameters and what logs you want collected after reproduction.
Why this is completely unacceptable from user perspective?
This is a regression that breaks a basic, real-world workflow (HDMI output to a TV) on a laptop that is still perfectly capable and widely used.
The part that feels almost insulting is the silence from NVIDIA: without actionable feedback, it looks like regressions on slightly older laptops are simply not a priority. Users end up pinned to 580.82.09-1 indefinitely, which is not a sustainable solution.
don’t waste your time trying newer versions, it’s still not fixed. 590 release feedback & discussion - #142 by aplattner
Just to clarify what I meant.
On any rolling-release Linux distribution, following normal updates is the expected workflow, and staying on an old driver actually requires extra effort (pinning packages, blocking updates). When a driver reaches the main repositories, users naturally assume it is intended for general use, not as an experimental build.
In this case, there is a consistent regression: HDMI at 120 Hz works on 580.82.09-1, breaks starting with 580.95.05-1, and remains broken in later releases, including nvidia-open.
I’m not intentionally testing newer versions — I’m simply keeping a rolling-release system up to date. From a user perspective, it is quite frustrating that this regression remains unfixed while affected driver versions continue to be published to the main repositories.
If this issue is already tracked internally as bug 5622970, I’d appreciate any guidance on specific diagnostics or testing that could help move this toward a resolution.