470 series driver on GTX 860M disconnects/forgets the internal laptop display when X is started

When I install the recommended driver for my GTX 860M (470.129.06) the internal laptop display works upon boot, but is disconnected/forgotten when X starts. It is also not listed when I run nvidia-settings or list the displays with xrandr.

I’ve recently upgraded my laptop (hdd → ssd) and reinstalled the OS (Debian 11.4). The previous installation was still using driver version 465.31 and this worked without any issues (and still does on this fresh installation).

It took me a while to discover that connecting an external monitor does work. Other than the ‘forgotten’ internal display (and PowerMizer thinking its always on battery power source) the driver seems to work just fine.

The output of nvidia-bug-report.sh confirms that the display is present at DFP-3 when using the 465.31 driver, but mysteriously missing/disconnected when using any driver from the 470 branch…

I’ve tried to use the “ConnectedMonitor” + “CustomEDID” X Config options to try and force it.

The driver then ‘sees’ the display and its backlight turns on, but is still unable to display anything (the display stays black). The display is now also available in nvidia-settings but the ‘Display Device Information’ tab shows that the connection link is “Disabled” instead of the “4 lanes @ 1.62 Gbps” that I normally see. Also the ‘DisplayPort Connector Type’ is listed as “Unknown” instead of “DisplayPort”.

Please help, the old 465.31 driver is bound to stop working at some point in the future and then I will have to resort to using Nouveau…

The laptop is an Alienware 17 (2015 model) with a GTX 860M and a 1080p@120Hz 3D Vision display. It has no integrated graphics, only the Nvidia card. The integrated graphics are disabled on this laptop due to the 120Hz refresh rate of the display and it is connected directly to the Nvidia display adapter. Atleast thats how Alienware explained it when I ordered the laptop in this configuration.

On a side note; I can make this laptop perform a thermal shutdown by booting into resque mode (ie. single user mode) and then running the Nvidia driver installer without specifing ‘-j1’.
Thats the only time that happens, so I try not to forget…
My bad for letting them put a i7-4910MQ in there, lol.

Edit: Added nvidia-bug-report.log.gz while running 470.141.03
nvidia-bug-report.log.gz (363.1 KB)

Please run nvidia-bug-report.sh as root and attach the resulting nvidia-bug-report.log.gz file to your post.

I’ve added the requested file to my post, thank you for being willing to take a look at it.

This post includes a nvidia-bug-report.log.gz while running driver version 465.31. The only real difference I’ve noticed so far is that the laptop display is detected/logged by this version and everything works.

Both reports were generated by first booting normally and then switching to a vt using ctrl-alt-F2 and running:

$ sudo systemctl stop sddm
$ startx -- -logverbose 6

Followed by:
$ sudo nvidia-bug-report.sh

nvidia-bug-report.log.gz (401.9 KB)

There’s really nothing logged that would point to the cause. It’s just either connected or disconnected.

That is also my conclusion. Wierd isn’t it? This is clearly a problem with the driver itself.
Any suggestions? Perhaps I should try to send the bugreport to NVIDIA via email.

Yes, absolutely. Please send the logs to linux-bugs[at]nvidia.com

I wil do that. Thank you for your help generix!

Is it customary NVIDIA does’nt send you any reply when you email them? I find this kind of rude…

IDK, I guess milage varies. To be honest, I suspected your chances to be low due to the 470 driver and Kepler now being legacy and in security maintenance only.
A skeleton goes to a doctor. “Now you should have been here earlier”.

I understand what you are saying. But there’s not even an automated reply to let you know that they have received the email. IMHO a bug with this level of severity should be fixed, no matter the legacy status.

Edit, re-worded my post to sound less harsh. (I was quoting some EU regulation that protects against hardware becoming e-waste too soon.). I’ve reconsidered and don’t want to be a ‘Karen’, I’ve fully enjoyed using this laptop/videocard for a long time :)

I am aware that this is an older card (released in 2014) and that writing drivers is hard. But this issue affects the basic operation of the device, ie. outputting a signal to a display. If it cant even do that…

I was hoping to get a couple of more years out of this machine. Currently I am on a 10-year upgrade cycle, meaning I build/buy the best system I can afford and that will have to do apart from some minor upgrades along the way.

I guess all I can can do is hope that the next security release contains a little surprise then.

I have tried Debian ‘testing/bookworm’ over the last couple of days and the 465 driver does not work for that :(
The 470 driver also has the same issue of ‘disconnecting’ the laptop display.

Thankfully the debian package for the 390 driver does work (but with a reduced level of OpenGL/Vulkan support).