Nvidia update broke my DVI to HDMI cable setup

Ask Fedora
1

Hi,

I have a desktop pc with fedora 40 that connects to the pc monitor using a DVI to DVI cable. I also have a DVI (PC) to HDMI (TV) cable to connect the pc to the tv. Up until yesterday it worked fine, but when I updated xorg-x11-drv-nvidia from 550.90 to 555.58.02-1 it stopped working. Everything else works fine, except for that DVI to HDMI connection. I was hoping somebody could give me an idea of how to solve it.

edit: I tried reverting the changes, but it says

Cannot find rpm nevra “xorg-x11-drv-nvidia-…-55.90”

for all the packages it updated yesterday

2

Do you know if the cable is “single” or “dual” link and whether it provides DDC?
How long is it?

“Smart” cables sometimes need a reboot. If the cable is always connected, try unplugging the cable and reconnecting. Newer drivers are sometimes intentionally made more fussy about strict adherence to standards – a web search for the cable model may show if others are encountering issues.

You find some error messages with journalctl --follow while hot plugging the cable. If that fails, try journalctl --no-hostname -b -g EDID| cat (the |cat wraps long lines) so the text is suitable for pasting into a forum post.

3

I tried fully disconnecting the cable and shutting both PC and TV down, but still doesn’t work. After doing journalctl --folow I get

Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): connected
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): Internal TMDS
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): 600.0 MHz maximum pixel clock
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0):
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): connected
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): Internal TMDS
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (--) 
NVIDIA(GPU-0): LG Electronics LG TV (DFP-0): 600.0 MHz maximum pixel clock
ul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (II) 
NVIDIA(0): Setting mode "DVI-D-0: nvidia-auto-select @3840x2160 +0+0 
{ViewPortIn=3840x2160, ViewPortOut=3840x2160+0+0}"
Jul 07 23:28:17 fedora /usr/libexec/gdm-x-session[2131]: (WW) 
NVIDIA(0): Failed to request fliplock.
Jul 07 23:28:17 fedora gsd-media-keys[2500]: gvc_mixer_card_get_index: 
assertion 'GVC_IS_MIXER_CARD (card)' failed
Jul 07 23:28:18 fedora gnome-shell[2373]: Can't update stage views actor 
unnamed [MetaWindowActorX11] is on because it needs an allocation.
Jul 07 23:28:18 fedora gnome-shell[2373]: Can't update stage views actor 
unnamed [MetaSurfaceActorX11] is on because it needs an allocation.

There are also logs related to DFP-1 through DFP-6 but they only state that they are disconnected

4

I have the same issue with a monitor connected with a DVI-to-HDMI-Adapter after updating to the newer nvidia driver (see Monitor stays black after update to Nvidia driver 555.58 | F40). It seems to be a common issue with this driver version (Bug :nvidia-drivers-555.58 DVI output not working ->black screen - Linux - NVIDIA Developer Forums)

5

Else where on this forum someone reported a bug raised against NVIDIA for breaking DVI to HDMI cables. Note the bug is against NVIDIA not rpmfusion.

You could downgrade to the 550 until the NVIDIA bug is fixed.

1 Like
6

There is no point in reporting most issues against nvidia at rpmfusion, we can’t fix issues in the closed source libs.
Anything beyond nvidia module compile issues or packaging issues needs to be directly reported to nvidia so they can fix it.

7

Sorry if my note was not clear, that is exactly what I wished to state.
Don’t report to rpmfusion.

8

Downgrade to 550 is possible, it’s a one line command (this forums formatting sucks).

sudo dnf --enablerepo rpmfusion-nonfree-nvidia-driver downgrade *\nvidia\* --allowerasing

9

Thanks for the one liner! I guess my issue was more not being able to revert the changes rather than the bug itself. I should have probably articulated it better. I was able to go back to 550 and it now works again. Thanks! :smiley:

1 Like
10

You may get important details for a bug report on Nvidia’s site by comparing journalctl messages for the two kernels. This is an example where a “workaround” tag would be preferable to “solution”. You are in a good position to provide the details required to solve the issue, so submitting a bug report may lead to a quick solution.

One test that may prove useful is to check that the problem occurs for a newly created user account.