Strategy for someone with an AMD APU + Nvidia GPU

[Hellothere]

In my last post (which I made a few moments ago) I was asking about nvidia gpu temps, but I’ve just remembered I have an AMD APU (Ryzen 5 3400G).
So a couple of questions:

1. If at all possible, how do I enable the igpu (I suppose there are better places to ask this, not here). My motherboard is an Asrock B450M steel legend.

2. Even if I did manage to enable my APU’s igpu, how would I use it in qubes (in an ideal situation I’d have qubes completely power down my nvidia card)?

[apparatus]

@Hellothere

Even if I did manage to enable my APU’s igpu, how would I use it in qubes

You can passthrough it to one of your qubes to use it there.

@Hellothere

in an ideal situation I’d have qubes completely power down my nvidia card

You can power it down like this:
https://wiki.archlinux.org/title/Hybrid_graphics#Fully_power_down_discrete_GPU

[ymy]

i think the iommu post doesnt apply here, @apparatus ?

[apparatus]

You’re right, my bad, I didn’t copy/paste correctly and didn’t check what was pasted.
The correct link is this one:
https://wiki.archlinux.org/title/Hybrid_graphics#Fully_power_down_discrete_GPU

[Hellothere]

Does that article work for dekstops as well? I was under the assumption bumblebee/prime/whatever is only supported on laptop cards

[apparatus]

@Hellothere

Does that article work for dekstops as well?

Yes.

[Hellothere]

Ok so I managed to activate my iGPU, and here is the output of inxi -G:
raphics: Device-1: NVIDIA GA106 [GeForce RTX 3060 Lite Hash Rate] driver: nouveau v: kernel Device-2: AMD Picasso/Raven 2 [Radeon Vega Series / Radeon Mobile Series] driver: amdgpu v: kernel Display: x11 server: X.Org v: 1.20.14 driver: X: loaded: modesetting unloaded: fbdev,vesa dri: radeonsi gpu: amdgpu resolution: 1920x1080~60Hz API: OpenGL v: 4.6 vendor: amd mesa v: 23.0.3 renderer: AMD Radeon Vega 11 Graphics (raven LLVM 15.0.7 DRM 3.57 6.10.3-1.qubes.25.fc37.x86_64)

Do I just follow the udev rules on the archwiki page, or is there something else I have to do?

Also why would using the udev rules be ‘unsafe’, because the wiki page has the line “With an NVIDIA GPU, this can be more safely done using bbswitch, which consists of a kernel package that automatically issues the correct ACPI calls to disable the discrete GPU when not needed, or automatically at boot.”

[apparatus]

@Hellothere

Do I just follow the udev rules on the archwiki page

Yes.

@Hellothere

Also why would using the udev rules be ‘unsafe’, because the wiki page has the line “With an NVIDIA GPU, this can be more safely done using bbswitch, which consists of a kernel package that automatically issues the correct ACPI calls to disable the discrete GPU when not needed, or automatically at boot.”

I don’t know.

[Hellothere]

So I followed the page, and indeed the nvidia card shows up nowhere on my system, but it still lights up. Is there a way to know if it’s only the RGB lights that light up, or if the GPU circuit is powered up as well?

[apparatus]

I guess the only way is to check the PC’s power consumption using wattage meter.

[Hellothere]

Ok so I had a chat on the nouveau IRC channel, and a dev (I think) said it shouldn’t work for desktops, and it’s just better to let the nouveau driver load so that it can power down the card to a sufficiently low power state (but it doesn’t turn it off completely). Apparently nouveau is mature enough that it can do this.

Also if you do want to use the card for anything you can do so by running DRI_PRIME=1 in the shell, or, if you have mesa 24.2 or newer, you can use zink as well: NOUVEAU_USE_ZINK=1 DRI_PRIME=1 <application>

You can find the chatlog here: IRC Logs of #nouveau on irc.freenode.net for 2024-08-13

[apparatus]

Thanks for the info, good to know. But I guess it’d be best to check it using wattage meter to make sure how it works.