I want to switch to Endeavour after using Manjaro for more than a year. How hard is it to keep Endeavour (or just Arch for that matter) stable with a legacy Nvidia GPU? The reason why I am concerned is because I saw some Internet “expert” telling someone to avoid using Arch with an older GPU. Manjaro has legacy drivers in their official repos and I expect AUR to work here just fine.
What I am more concerned about is that Manjaro also has specific packages combining the kernel version and a driver version (i.e. linux61-nvidia-470xx). They are nowhere to be found on Arch. Do I have to compile them myself with each update or does it just work without that? If so, I’m kinda curious why Manjaro does it this way.
Never used Manjaro, so I can’t speak to that, but they probably did it to try and provide support for older hardware. I also have no experience with legacy drivers, but I do have experience using dkms versions of drivers to have them compiled into each new kernel.
I believe the package you are looking for is https://aur.archlinux.org/packages/nvidia-470xx-dkms.
As for how reliable the legacy dkms driver is, I can not say. However, I have had no problems with nvidia-dkms, so far with the current proprietary driver.
Manjaro also has nvidia-470xx-dkms so I’m not sure. Maybe it’s my mistake for having them both installed? I just assumed everything with “470xx” in the name is needed for the driver to function
Yeah you probably didn’t need both. Normally you either use the driver that matches the kernel or you use dkms version of driver to have it compiled into the kernel.
I used Arch/EOs with a very old NVIDIA GPU for some years (390xx drivers), normally the AUR packagers include, when needed, a patch for newer kernel version, so it should be ok-ish (at least on XOrg).
Just pay attention to:
initial driver choice (dkms vs kernel-specific)
updates:
always wait a couple of days and check AUR comments relative to your driver, normally problems will arise there
always read the output of init regeneration (mkinitcpio/dracut…) for eventual issues to be solved before rebooting
in the event of a driver failure (typical black screen on startup), repress the urge of a long sequence of blasphemies and boot up a live iso, chrooting on your system and attempting to downgrade (if you’re on BTRFS with a sane snapshot workflow, you can attempt a much easier rollback).
That’s at least my experience (rather patchy and unprofessional, naturally) with an old NVIDIA card…it’s manageable and you can leverage an old-still-good card that way…hope it helped
Thanks for replying. Yeah, I was planning to use BTRFS. And thanks for letting me know about mkinitcpio, didn’t know about that, I’m sure it’s going to be useful knowledge
glad to have contributed, @qunas…
as far as I remember, the issues in the init regeneration popped up when using the -dkms version - this one in your case, as mentioned by @d-air1 .
More in general, though, pay attention to the drivers building process, as it usually leaves a log detailing the eventual error from which you can start to elaborate a solution.
good luck
