You are not logged in.
For 390xx, I had to repeat the manual patching again, after the last update few days ago, for PRIME synchronization to work. I am using the latest kernel.
Offline
My measure of "doesn't work" is that there is still tearing. I use the conf that worked before the whole shitstorm, although every time I make an update, I have to check whether the system uses eDP1, eDP-1 or eDP-1-1 for my screen name, change my conf files accordingly, and hope there isn't a new fancy option that has to be enabled in order to not get a black screen.
At this point I don't even remember if I originally had the "xrandr --setprovideroutputsource" thing uncommented in my xinitrc (I mean, before switching to LTS), but it fails with the same errors as it was reported above by other people (I have Intel and NVIDIA-G0 providers).
Then I decided to drop this and use NVIDIA only, so I removed xf86-video-intel, changed my screen name again in my conf files (whoever is messing around with the name schemes, fuck you), now I have modesetting and NVIDIA-G0 as providers (but still no luck with xrandr otherwise), still tearing, applications are rendered through MESA, but *somehow* prime-run manages to run them on my discrete GPU (I can confirm that in nvidia-smi or when said application shows me the vendor string), still tearing.
What's left to try besides rolling back to linux-lts 4.19 from last month ?
Offline
(I have Intel and NVIDIA-G0 providers).
What's left to try besides rolling back to linux-lts 4.19 from last month ?
prime and render-offload prime are not the same despite the terrible idea of naming them both prime. and from what your describing it sounds like your using a render-offload prime setup otherwise you wouldnt even have `prime-run` installed.
this thread is related to regular PRIME and prime sync.
Offline
I have prime-run installed as a desperate measure to find something that works, it's just a script to wrap the execution with a few environment variables anyway, so I figured why not give it a try.
edit : ok actually there was an extra conf file with the package so I removed it, but now I either have modesetting as a provider when I try to configure with NVIDIA alone, either Intel as a provider when I reinstall xf86-video-intel.
In both cases, I still have tearing and can't use my GPU anyway.
edit 2 : nevermind, "chmod 666 /etc/X11/xorg.conf" fixed the problem. ;-(
Last edited by lsbkf (2020-02-29 11:20:57)
Offline
I am also still seeing this with linux-5.5.13.arch2-1 and nvidia-dkms-440.64-9.
~ xrandr --output eDP-1-1 --set "PRIME Synchronization" 1
~ xrandr --verbose | grep PRIME
PRIME Synchronization: 0
Also with the non dkms, it seems like the patch no longer applies successfully for a manual build based on earlier instructions.
Offline
Also took a crack and using asp and makepkg to incorporate this patch: https://gitlab.com/snippets/1925454
The issue is still present.
Offline
sudo pacman -U https://archive.archlinux.org/packages/ … pkg.tar.xz https://archive.archlinux.org/packages/ … pkg.tar.xz https://archive.archlinux.org/packages/ … pkg.tar.xz https://archive.archlinux.org/packages/ … pkg.tar.xz
Attempted a kernel downgrade, issue still present.
Offline
Please open up your own thread and post details of your configuration (xorg.log, what you did to configure the nvidia gpu, any config files created as a result). The issue as discussed here is fixed with 5.5 and the latest nvidia driver, it's likely you misconfigured something to not meet the necessary preconditions.
Offline
hallo i run kernel 5.8.10 kernel and nvida dkms 450.66 had also test with 455.23.04 installed. see that xrandr --prop show my that prime synchronization is 0 on my eDP-1-1 output.
any had experience and can to confirm.
thanks all
Offline
Closing this old thread.
If this is in any way still present, open your own thread post new details and how you've configured PRIME Synchronisation.
Offline