Nvidia graphics driver naming convention
Nicolas Marie
Hello.
I've got three workstations using different Nvidia GPU. I upgraded them from Debian 11.6 to Debian 12:
- The first one is a Geforce GT520 (Fermi) so I used the package nvidia-graphics-drivers-legacy-390xx and it works like a charm.
- The second one is a Geforce 1050 Ti (Pascal) si I used nvidia-graphics-drivers which is the version 525 and it also works like a charm.
- The last one is a Geforce GT730 (Kepler) so I have to use the Nvidia driver version 470.
There isn't any package named nvidia-graphics-drivers-legacy-470xx, like we can find for the 390 mentioned version or the older 340 version. As a workaround, I used nvidia-graphics-drivers-tesla-470 and all the GLX stack / SMI utility / nvidia-settings utility. Basically, all seems to be working:
- The nvidia module is built through the DKMS workflow and loaded.
- The GPU is recognized.
- The basic graphics acceleration (KDE windows effects / GLXGears) works without any glitch.
My question is about the package naming: is the nvidia-graphics-drivers-tesla-470 specifically built for Tesla graphics, and the fact that I can use it for my consumer Nvidia GPU is a positive "side effect"? Or is this package totally usable with a consumer GPU and, in this case the name nvidia-graphics-drivers-legacy-470xx would be more relevant for the non-aware debian new users. As a comparison, Ubuntu uses a similar naming convention, namely nvidia-driver-390, nvidia-driver-470 and nvidia-driver-525.
P.S.: I know that this question type should be escalated during
beta states but, since I used the Debian Bullseye Backports
packages for nearly one year, I wrongfully believed these drivers
were be maintained in the next Debian release.
Thanks for your answer.
Regards.
Andrew M.A. Cater
> Hello.
>
> I've got three workstations using different Nvidia GPU. I upgraded them from
> Debian 11.6 to Debian 12:
>
> *nvidia-graphics-drivers-legacy-390xx* and it works like a charm.
> * The second one is a Geforce 1050 Ti (Pascal) si I used
> *nvidia-graphics-drivers* which is the version 525 and it also works
> like a charm.
> * The last one is a Geforce GT730 (Kepler) so I have to use the Nvidia
> driver version *470*.
>
https://wiki.debian.org/NvidiaGraphicsDrivers gives the canonical instructions
- I think that one's old enough that nouveau also works well with it too.
> There isn't any package named *nvidia-graphics-drivers-legacy-470xx*, like
>
> * The GPU is recognized.
> * The basic graphics acceleration (KDE windows effects / GLXGears)
> works without any glitch.
>
See above.
> P.S.: I know that this question type should be escalated during beta states
> but, since I used the Debian Bullseye Backports packages for nearly one
> year, I wrongfully believed these drivers were be maintained in the next
> Debian release.
>
> Thanks for your answer.
> *
My standard advice also is to build Nvidia drivers from a bare CLI text only install *before* going back to install a desktop environment and preferably using
the expert text mode install, but that may well be overkill.
It does, however, prevent nouveau or any other conflicting packages causing
problems if you are determined to use hte proprietary drivers.
As ever, your mileage may vary.
With every good wish,
Andy Cater
Anssi Saari
> My question is about the package naming: is the nvidia-graphics-drivers-tesla-470 specifically built for Tesla
> graphics, and the fact that I can use it for my consumer Nvidia GPU is
> a positive "side effect"?
GPU families are supported and gives a reference to a documentation file
with more details.
I have no idea about the why of the tesla naming. I'd guess it's a
historical reason and lack of manpower to update.