Gerardo Zamudio,
The SBo release does not yet target current. I have to move some stuff around to take into account that libglvnd is in Slackware-current. The symlinks will be adjusted appropriately since with libglvnd we can now have multiple drivers installed without everybody stomping on everybody else. I will not do that until RC1 hits and the final kernel is locked down. If you want to use Nvidia SBo on current, some issues may arise until I can sort everything out.

see also: https://github.com/NVIDIA/libglvnd

That said, I will most likely update for 14.2 since I have run into some display port issues.

FWIW I just installed the new mesa-19.1.0 (in testing), built nvidia-430.26 driver (Slack64-current w/ Plasma 5) and so far everything is running just fine.

3 members found this post helpful.

Ditto, but using Xfce and Nvidia's 430.26 blob.

1 members found this post helpful.

I took the opportunity to switch over to the blob version. Seems to have gone swimmingly.

1 members found this post helpful.

Since there are relatively frequent kernel updates I just want to be clear on what to do (when running the Nvidia blob) after building and booting the new kernel. It looks like -

nvidia-switch --remove

build and reinstall nvidia-kernel and nvidia-driver

reinstall mesa

startx

If I have missed something or one or more of these steps are not needed please let me know. I don't want to mess this up. Thanks in advance.

Shouldn't mesa be installed before installing the Nvidia blob?

Not sure of the proper sequence hence the question.

this is the tut i used and i have the nvidia drive up and running pergectly


https://linuxconfig.org/install-and-...ulkan-on-linux

I just run 'sh NVIDIA-Linux-x86_64-430.26.run' after booting the new kernel or after a mesa upgrade.

The installer removes the installed driver and installs everything needed I think.

I don't reinstall mesa.

I think the SBo NVidia package is usefull when you have more than one kernel installed.
NVidia blob needs to be reinstalled every time you boot a different kernel.

OK, thanks for the replies. I'll reinstall with each kernel upgrade. Thanks again

Just FTR that's how I install nVidia,too. It takes less than 5 minutes to let the .run file do it's work and always works great. However the need for reinstallation appears to be diminishing in importance. I have successfully tried merely modprobing nvidia where versioning was chosen to be ignored 3 out of 5 last kernel upgrades. Currently it is inconsistent and rather a rolling of the dice, but even that 5 minutes may be avoidable, hopefully even moreso as variations are reduced.

You don't have to rebuild the nvidia-driver package every time the kernel is upgraded just rebuild the nvidia-kernel package. If you are using the .run file you can specify the -K advanced option to just build the kernel module if the driver is already installed e.g:
I only rebuild the driver if the mesa and/or xorg-server packages are upgraded.

3 members found this post helpful.

Thanks, you got the correct answer up before me. This saved me from yelling pounding on the screen yelling "NEIN! NEIN! NEIN!"
https://www.youtube.com/watch?v=xoMgnJDXd3k

2 members found this post helpful.

Then if I have two kernel versions (I have the stock -current kernel and 5.1.x) I can first install the, let's say,
"main entire" NVidia blob by running "sh NVIDIA-Linux-x86_64-430.26.run" under any kernel,
and then I can put "sh NVIDIA-Linux-x86_64-430.26.run -K $(uname -r)" in /etc/rc.d/rc.local, this way only
the kernel driver will be built every boot for that particular kernel version.
Will this work?


edit-
This could be better, just build the driver if it doesn't exist.

From my understanding, this is basically what dkms will do automatically if you have it installed with programs that utilize it (I believe at least Nvidia and VirtualBox support dkms). Make sure you read dkms's README.

1 members found this post helpful.