let’s discuss it here 
Yes, nvidia-inst seems to operate incorrectly with those options.
Just noticed that instead of options -top if only -to is given, then it seems to produce correct output. So option --prime causes an error here.
Will look into it as soon as possible.
1 Like
question is if we may need to implement something to automatically ask in case user have very new GPU to use the nvidia-open modules as of they are recommended by Nvidia for future improvements.
1 Like
That sounds like a good idea.
a bit confusing but it shows this:
isn’t NV190 family (Ada Lovelace) newer compared to NV160 family (Turing) ?
on the first line , they should add ada lovelace (NV190/GKXXX) for Nvidia-Open
yes thats what i think the same but its added on second line..
I can not find the info on nvidia anymore.. remember someone posted that before ..
or for trouble on last drivers window$ , they work less a time under linux drivers for theses cards at the moment
Not released yet, but I’ve fixed the problem with the --prime option.
Then I added a check for Turing and newer GPUs to automatically install the open source Nvidia driver for these newer GPUs. This automation can be bypassed with either of the options --open or --closed.
But this automation is experimental now, and needs to be tested with Turing and newer GPUs. I have only an older Nvidia GPU, so I can’t test if everything works right away.
The old behavior (without automatic selection) is achieved with command
nvidia-inst --closed
Edit: just released nvidia-inst 25.4.1-1. Will be in all mirrors within a few hours, some mirrors have it much faster.
Yes. The time order from oldest to newest:
- maxwell
- pascal
- volta
- turing
- ampere
- ada lovelace
- blackwell
Maxwell is the first supported by nvidia-dkms.
Turing is the first supported by nvidia-open-dkms.