535.154.05 low and heavily fluctuating fps, occasional black screen

Graphics / Linux Linux
1

Hi there,
here are my issues with this driver version.

#1. Since installing this 535 series I have occasional black screens for 1-3 seconds when e.g. switching active windows. Starting steam is guaranteed to cause this.
Maybe this is related to?

[drm:nv_drm_master_set [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership

#2. This is the most pressing issue and actually unbearable!
I have set the refresh rate to 144Hz to match my displays specs and this driver is not able to provide a steady framerate in plain desktop mode. See attached screenshot.
Neddless to say, that this is happening in games also and makes me seasick!

Bildschirmfoto vom 2024-01-30 11-19-28
Bildschirmfoto vom 2024-01-30 11-19-281098×595 137 KB

Please find attached the bug report. As far as I can see, all specs are included.
nvidia-bug-report.log.gz (579.5 KB)

2

This rather sounds like a compositor/cinnamon issue. Which driver version was installed before?
For (fullscreen) games, cinnamon should have an option to turn off the compositor, is that enabled?

3

I had the 525.147.05 driver running with none of these issues before.

I just looked up on the “compositor” - never heard of before…
I checked the setting and is not “deactivated for fullscreen”.

While this might help for fullscreen apps (will test tonight), is has no effect on the desktop frame rate. I just enabled the option and restarted cinnamon.
Additionally, from what I read, deactivating the compositor might lead to tearing.

4

For a bad workaround, you could try if enabling forcecompositionpipeline enhances desktop smoothness, please create
/etc/X11/xorg.conf.d/nvidia-comp.conf

Section "OutputClass"
    Identifier "nvidia-composition"
    MatchDriver "nvidia-drm"
    Driver "nvidia"
    Option "ForceCompositionPipeline" "true"
EndSection

For gaming, usually all composition should be disabled.
Nevertheless, I think this is rather a cinnamon issue, surfacing with driver change being a coincidence. Maybe also attach here:
https://github.com/linuxmint/cinnamon/issues/11957

5

While I was gathering the information regarding my previously installed driver, that you asked, I noticed the there was also a 545.29.06 driver available.

Rather than trying “bad workarounds” I decided to give that a try first.
I pulled a timeshift snapshot and went for it.
And here you have it.
There are still some dips but I guess that is not noticeable.

Bildschirmfoto vom 2024-01-30 18-48-56
Bildschirmfoto vom 2024-01-30 18-48-561345×765 181 KB

I’m curious if the “disabled compositor” will add more benefit in games and I might also give the “bad hack” of yours a try. XD

I’m attaching the new bug report for completeness and I guess this ticket can be closed.
Thanks for your assistance @generix!

nvidia-bug-report.log.gz (612.6 KB)

Edit:
I forgot about the black screens…also went away.
Although the kernel log is still showing those entries.

[drm:nv_drm_master_set [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
6

I have multiple errors from nvidia_drm driver at startup.

Configuration:

  • cat /etc/lsb-release
    DISTRIB_ID=Ubuntu
    DISTRIB_RELEASE=23.10
    DISTRIB_CODENAME=mantic
    DISTRIB_DESCRIPTION=“Ubuntu 23.10”

  • Linux ******** 6.5.0-15-generic #15-Ubuntu SMP PREEMPT_DYNAMIC Tue Jan 9 17:03:36 UTC 2024 x86_64 x86_64 x86_64 GNU/Linux

  • GPU RTX 3070ti

modinfo nvidia_drm

filename: /lib/modules/6.5.0-15-generic/kernel/nvidia-535/nvidia-drm.ko
version: 535.154.05
supported: external
license: MIT
srcversion: 2A9024C4A3362B9F6696FDD
alias: pci:v000010DEdsvsdbc06sc80i00
alias: pci:v000010DEdsvsdbc03sc02i00
alias: pci:v000010DEdsvsdbc03sc00i00
depends: drm,drm_kms_helper,nvidia-modeset
retpoline: Y
name: nvidia_drm
vermagic: 6.5.0-15-generic SMP preempt mod_unload modversions
sig_id: PKCS#7
signer: Canonical Ltd. Kernel Module Signing
sig_key: …
sig_hashalgo: sha512
signature: …
parm: modeset:Enable atomic kernel modesetting (1 = enable, 0 = disable (default)) (bool)

dmesg errors at startup:

[ 23.397441] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 29.914782] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 30.869013] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.164981] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.165054] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.165083] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.165115] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.280541] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.280592] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.280612] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.280629] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.333198] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.333287] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.333312] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership
[ 34.333334] [drm:nv_drm_master_set [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to grab modeset ownership

7

can be safely ignored.

8

Thank you Generix

It is what I though because I tried glxgears benchmark and I didn’t notice a GPU processing derating, 144Hz steady.

Do you know if such benign error is also present with the nvidia drivers closed source version ?
I don’t know if It’s better to keep installed the open one…

9

The error is not triggered by the nvidia driver but by a misbehaving application you’re running. So this won’t change with driver versions.