So, if I have 2 GPU's in my system, an AMD (primary) and an nvidia (secondary, not hooked up to an o

So, if I have 2 GPU's in my system, an AMD (primary) and an nvidia (secondary, not hooked up to an output, just for AI/ML), what's the cleanest way to get the nvidia GPU working with the 'bazzite-gnome' image? I did not install the nvidia one, as my primary GPU for gaming is an AMD.

Aanoldguy So, if I have 2 GPU's in my system, an AMD (primary) and an nvidia (secondary, n...

antheas•2/25/24, 5:41 PM

i will wager a guess and say nvidia-docker

antheas•2/25/24, 5:42 PM

im not 100% sure but i think you dont need an nvidia driver on your computer to use it with nvidia-docker

anoldguyOP•2/25/24, 5:42 PM

Ooo, I have done zero nvidia-docker. I'll poke it. TIL. Thanks for the nudge.

©TriMoon™•2/25/24, 5:42 PM

Would be cool to be able to game on the nVidia in a setup like that, cause im planning to get rid of my nVidia in long run...

antheas•2/25/24, 5:42 PM

its just a docker extension, might have gotten renamed too

antheas•2/25/24, 5:43 PM

works well. But every time I used it I had some sort of nvidia driver installed

antheas•2/25/24, 5:43 PM

if you use a cuda image though the cuda is in the image, maybe that extends to the driver too

©TriMoon™•2/25/24, 5:43 PM

So no need for any nVidia crap on the host at all? just inside a docker? wow

©TriMoon™•2/25/24, 5:44 PM

I will need help with setting that all up then, when i reinstall Bazzite

anoldguyOP•2/25/24, 5:44 PM

Trying now. Gotta get my tabby running again. https://tabby.tabbyml.com/

Home | Tabby

Description will go into a meta tag in

anoldguyOP•2/25/24, 5:45 PM

Wow, great embed, folks. (Lets you run LLM locally)

©TriMoon™•2/25/24, 5:45 PM

LLM

(Acronym not found in db error...)

antheas•2/25/24, 5:46 PM

large language model

©TriMoon™•2/25/24, 5:46 PM

ahh

antheas•2/25/24, 5:46 PM

i tried to make one that helped you talk to women, never worked that well

anoldguyOP•2/25/24, 5:46 PM

Oof.

antheas•2/25/24, 5:46 PM

but it had some moments

antheas•2/25/24, 5:47 PM

said a lot of racist stuff

©TriMoon™•2/25/24, 5:47 PM

An AI is like a kid, you get out from it what you put in to it

anoldguyOP•2/25/24, 5:49 PM

https://cloud.githubusercontent.com/assets/3028125/12213714/5b208976-b632-11e5-8406-38d379ec46aa.png Well, crap. Looks like it still requires a cuda driver on the host OS.

©TriMoon™•2/25/24, 5:49 PM

bleh

anoldguyOP•2/25/24, 5:50 PM

©TriMoon™•2/25/24, 5:50 PM

Nothing new, we didnt expect any better from nVidia did we?

anoldguyOP•2/25/24, 5:50 PM

No. I did not.

©TriMoon™•2/25/24, 5:51 PM

Time to be freed from the nVidia virus soon for me (I hope)

antheas•2/25/24, 5:52 PM

unfortunate

antheas•2/25/24, 5:52 PM

and i think nvidia has banned pcie passthrough for consumer cards

anoldguyOP•2/25/24, 5:52 PM

My issue is that even the 7900xtx is sloooooow for LLM/AI compared to a 3080. I'll take the FPS hit and lack of RTX for gaming, but...slow autosuggest AI in my editor?

I gotta draw the line somewhere.

©TriMoon™•2/25/24, 5:53 PM

Unfortunatly the iGPU even on my Rysen9 CPU aint strong enough to get rid of this stupid nVidia...

©TriMoon™•2/25/24, 5:54 PM

i got the 7950x3D... if i didnt mistype that models name

anoldguyOP•2/25/24, 5:54 PM

So, then the million-dollar question. Can I layer the nvidia driver on my current bazzite-gnome image?

I'll have to read more on ostree snapshots.

©TriMoon™•2/25/24, 5:55 PM

Nope i was correct:

©©TriMoon™i got the 7950x3D... if i didnt mistype that models name

anoldguyOP•2/25/24, 5:55 PM

I went with the non-x 7900. 65w cpu good enough for all my needs.

Aanoldguy I went with the non-x 7900. 65w cpu good enough for all my needs.

©TriMoon™•2/25/24, 5:56 PM

You probably did the right thing....
I hoped it would have been strong enough but...ahh well

Aanoldguy So, then the million-dollar question. Can I layer the nvidia driver on my curren...

©TriMoon™•2/25/24, 6:00 PM

Isn't there a Bazzite-nvidia-gnomeBazzite-nvidia-gnome version?

Or is that what SilverblueSilverblue actually is?

so many OCI variants lol

anoldguyOP•2/25/24, 6:03 PM

There is! But I'm not sure how that will play with my AMD card. And...I'm hesitant to rebase, never having done it before.

anoldguyOP•2/25/24, 6:04 PM

I'm def not the person to ask. I'm so new to this distro and approach. Loving it, but very new.

Niklas ⚡•2/25/24, 6:06 PM

At least tabby can be run on Apple silicon Macs

anoldguyOP•2/25/24, 6:08 PM

Oh yeah. I use it at work on my laptop. SO GOOD. And fast. I run both DeepseekCoder-6.7B and Mistral-7B. mods hits Mistral, and the VS Code plugin uses DeepseekCoder.

anoldguyOP•2/25/24, 6:10 PM

Ok, going to follow the instructions here for rebasing (https://universal-blue.org/images/), here's hoping for the best.

Glad I have a full backup, just in case.

LelouBil•2/25/24, 6:16 PM

Hey, on my system I can't log back in after sleep

LelouBil•2/25/24, 6:18 PM

and now I just get a message about the lock screen being broken

LelouBil•2/25/24, 6:18 PM

but ttys are just blinking cursors

anoldguyOP•2/25/24, 6:28 PM

Hmm. I had a 'staged' layer, so I attempted a reboot, and now I can't login to gnome. I can get a pty, and login on the console, but gdm hangs after I enter my password, no session. No interesting log entries in journalctl -xe -u gdmjournalctl -xe -u gdm except an entry for setroubleshootsetroubleshoot not being able to write to /var/lib/setroubleshoot/setroubleshoot_database.xml/var/lib/setroubleshoot/setroubleshoot_database.xml

anoldguyOP•2/25/24, 6:29 PM

Oh, hmm. There is a gnome-shellgnome-shell error: g_error_free: assertion 'error != NULL' failedg_error_free: assertion 'error != NULL' failed. But I have no context to know where that's coming from.

Aanoldguy Oh, hmm. There is a `gnome-shell` error: `g_error_free: assertion 'error != NULL...

Kyle Gospo•2/25/24, 6:37 PM

Try booting your previous image

Kyle Gospo•2/25/24, 6:37 PM

See it that works

Aanoldguy I went with the non-x 7900. 65w cpu good enough for all my needs.

Eeyore_Syndrome•2/25/24, 6:38 PM

7950x but only because I got a bundle deal at Micro Center last November lol.

And I didn't wanna deal with split P-Core issues.......

So, if I have 2 GPU's in my system, an AMD (primary) and an nvidia (secondary, not hooked up to an o

Similar Threads

Similar Threads

Similar Threads