Rgb glitch on the screen, newer kernel

Issues & Assistance
1

can someone please identify my issue?

img_lowres
img_lowres1080×597 109 KB

cachy os and gnome with gdm (installed kde at first, then switch immediately)
install date: 2024-07-05
specs: amd cpu 6800h with nvidia 3070ti
asus laptop using supergfxctl on Hybrid
display: 1080p 360hz (i don’t think this be the issue)
Nvidia drivers: current version NVIDIA-SMI 565.77, i noticed the issue on the nvidia package and moving to nvidia-open-dkms the issue still persists
display server: xorg, i have the issue on wayland too (i don’t use wayland cause i don’t get some refresh rate options)

the problem slowly crippled upon kernel updates. at first it was barely noticeable but nowadays its on its full glitch (every 1-2 seconds depending on graphical drawing)
on kernel 6.6.66 (cachyos lts) i have no issues but when i use the updated current kernel (6.12), zen kernel (6.12.4) or others i get this green screen glitch ( its sudden but goes away under milliseconds, sometimes red and green and…)
when any changes happen on the screen i get this sort of rgb glitch

what can be the cause of the issue? anything i miss? i only have it on non-lts kernel (6.10 above or something? i haven’t tested)

all the logs (can’t put individual links it seems):

ill be waiting for further questions and logs to send, to be honest this been boggling my head.

any kernel params to use in the build since last install?

thanks for the help!

3

ah forgot that one

/etc/mkinitcpio.conf
HOOKS=(base udev autodetect microcode modconf block keyboard keymap consolefont plymouth resume filesystems fsck)

modules, binaries and files got nothing

5

thanks for the quick reply

MODULES=(nvidia nvidia_modeset nvidia_uvm nvidia_drm)
HOOKS=(base udev autodetect microcode modconf block keyboard keymap consolefont plymouth resume filesystems fsck)

GRUB_DEFAULT='saved'
GRUB_TIMEOUT='5'
GRUB_DISTRIBUTOR='CachyOS'
GRUB_CMDLINE_LINUX_DEFAULT='nowatchdog nvme_load=YES zswap.enabled=0 quiet splash resume=UUID=e1da91c8-eaa6-4103-91dd-a9deaeea65a3 loglevel=3 amdgpu.sg_display=0 modeset=1 fbdev=1 nvidia_drm.modeset=1 amd-pstate=guided nvidia_drm.modeset=1 nvidia_drm.fbdev=1'
GRUB_CMDLINE_LINUX=""

so i added these and updated grub and did mkinitcpio -P, and reinstalling nvidia-open-dkms due to automatic hooks and dkms but i get this error:

==> ERROR: module not found: ‘nvidia’
==> ERROR: module not found: ‘nvidia_modeset’
==> ERROR: module not found: ‘nvidia_uvm’
==> ERROR: module not found: ‘nvidia_drm’
==> ERROR: module not found: ‘nvidia’
==> ERROR: module not found: ‘nvidia_modeset’
==> ERROR: module not found: ‘nvidia_uvm’
==> ERROR: module not found: ‘nvidia_drm’

modinfo (this is for nvidia) and others do return stuff:

❯ modinfo nvidia
filename:       /lib/modules/6.12.5-2-cachyos/updates/dkms/nvidia.ko.zst
firmware:       nvidia/565.77/gsp_ga10x.bin
firmware:       nvidia/565.77/gsp_tu10x.bin
license:        Dual MIT/GPL
supported:      external
version:        565.77
alias:          char-major-195-*
import_ns:      DMA_BUF
softdep:        pre: ecdh_generic,ecdsa_generic
name:           nvidia
depends:        
alias:          pci:v000010DEd*sv*sd*bc03sc00i00*
alias:          pci:v000010DEd*sv*sd*bc03sc02i00*
alias:          pci:v000010DEd*sv*sd*bc06sc80i00*
srcversion:     E352A918DD79D68C88AFA54
vermagic:       6.12.5-2-cachyos SMP preempt mod_unload 
retpoline:      Y
sig_id:         PKCS#7
signer:         DKMS module signing key

...
6

Is it glitching on the internal display? or the external one?

7

You don’t need any of those. It’s already patched in the driver. amd-pstate also has nothing to do with this.

1 Like
8

internal display (i don’t have any external connected atm)
i don’t get it on the lts kernel though

9

This is an ongoing issue with AMD displays and PSR. Adding amdgpu.dcdebugmask=0x10 to your boot parameters disables PSR and should mitigate this.

1 Like
10

woah, thank you!
been trying to find the solution thinking its on my end for the last 3 months.

checked and seems like this issue has persisted for a long time (even on fedora)
should we report this somewhere or? seems a critical issue.

anyway thanks for the help :green_heart: :green_heart:

11

It has already been reported, and there seems to be a patch that fixes this. I’ll build a kernel for you with the fix and check if it works.

1 Like
12

@Vphy Nginx Directory can you try this kernel? Make sure to remove amdgpu.dcdebugmask=0x10 when testing. Thanks.

1 Like
13

works flawlessly awesome!
would be great to apply this for future upstream

1 Like