I’m running Ubuntu 20.04 with XFCE using a new RTX 3060 card connected to an LG C1 OLED. I upgraded from CPU based graphics and the difference is huge. I’m leveraging the latest 510.54-0ubuntu0.20.04.1 package driver from the nvidia-driver-510. I was surprised by amount of weird manual configuration needed to get things stable such as manually needing to configure kernel modules, blacklisting Nouveau etc. One of odd things remaining is when I log out of X instead of going back to my login screen I get dropped to a monitor power saving screen. I have to use CTRL ALT F1 to switch to a text console and then CTRL ALT F6 or F7 to get back to an X login prompt. Anyone know why that happens and how I can get back to just having a login prompt appear again minus the power saving screen?
This continues to be an issue on Ubuntu 22.04 with the latest XFCE using driver 515.86.01-0ubuntu0.22.04.1.
Nothing but crickets for 9 months isn’t good enough. It’s truly sad it is impossible to get support from NVIDIA.
I attempted get support by escalating through support channels only to have a 2nd level support agent point me back to this forum that no one responds to and attempt to close the ticket. Perhaps some media attention will fix the problem…
Which DM is used on that?
I’ve tried a few to the same results. I’m using lightdm. It seems to be closest to functional of what I’ve tried.
Please run nvidia-bug-report.sh as root and attach the resulting nvidia-bug-report.log.gz file to your post.
nvidia-bug-report.log.gz (847.5 KB)
As requested.
Changes in misbehaviour of late. I still have the original problem but now frequently after login I’m in the wrong display resolution and I sitting with Display window showing 3840 x 1960 16:9 60 Hz instead of the 4096 x 2160 256:135 just below 120Hz. Setting it to the correct resolution works around the new issue but that should not be happening either.
The monitor is connected over hdmi 2.0 and uses FRL.
There are always moments where the driver loses connection to the monitor
dmesg:
nvidia-modeset: WARNING: GPU:0: HDMI FRL link training failed.
Xorg log:
[1265565.222] (--) NVIDIA(GPU-0): DFP-0: disconnected
[1265565.222] (--) NVIDIA(GPU-0): DFP-0: Internal TMDS
[1265565.222] (--) NVIDIA(GPU-0): DFP-0: 165.0 MHz maximum pixel clock
[1265565.222] (--) NVIDIA(GPU-0):
[1265571.605] (--) NVIDIA(GPU-0): LG Electronics LG TV SSCR (DFP-0): connected
[1265571.605] (--) NVIDIA(GPU-0): LG Electronics LG TV SSCR (DFP-0): Internal TMDS
[1265571.605] (--) NVIDIA(GPU-0): LG Electronics LG TV SSCR (DFP-0): 42666.7 MHz maximum pixel clock
Did you already try a different/shorter cable?
Please delete /etc/X11/xorg.conf though this doesn’t have anything to do with your issue, it has an intel gpu referenced which might get you into other trouble.
I did use other cables originally but they did not support the needed signalling. This one is a 2 meter
UGREEN HDMI 2.1 Cable 48Gbps Braided HDMI Cable specifically listed as supporting the 4K 120Hz mode I wanted. I can’t get much shorter than 2 meters and still connect the monitor. This is going to be really stupid if after all this it is the cable that is doing it. I deliberately tried to get a good quality cable originally. I’ll order another cable which lists “certified cable” and see if it helps.
I had assumed the FRL linking training messages were from when it attempted to do that when the monitor is powered off. Given the power usage of the panel I always turn it off when not in use. I will also try regenerating the xorg.conf to see if that helps.
I purged the xorg.conf configuration file. It is weird that the nvidia-settings tool cannot write a new xorg.conf even as root. I had to fix that with: chmod u+x /usr/share/screen-resolution-extra/nvidia-polkit
It is frustrating how cable manufactures can claim to support standards but do not work. I deliberately avoided cheap Chinese clones previously. My previous cable was twice the cheapest cable available and explicitly claimed to support the frequencies and signals I needed. I have replaced that cable with a brand new ultra certified one rated for the signals from Startech.com with gold connectors that was twice the price of the previous cable. I still do not believe there is anything wrong with the previous cable. It remains to be seen if that does anything but this action should fully eliminate a cable as a possible cause. I haven’t seen the display resolution glitch since replacing the configuration and fixing the missing permissions needed by nvidia-settings. I will try to get a bug filed against Ubuntu 22.04 bug to get nvidia-settings 510.47.03-0ubuntu1 fixed so other users are able to write their xorg.conf files. This was attempted by someone else in 2016 as bug #1625119 against the nvidia-settings package and that was not responded to. Granted that person was a bit snarky in their description and did not give a fix.
I posted Ubuntu papercut bug 2003004 about nvidia-settings not being able to write out configuration files as root.
Hopefully someone will fix it.
Issue persists with an expensive certified good cable. Other ideas?