Help to solve CPU peaks (overheats) while using camera (possibly HW Acceleration I guess)

[ksvivek]

Thanks for inspiring

[ksvivek]

I am trying the first method of trying nvidia , but stuck in this issue:
Created separate thread to avoid cluttering here

Error persists even after allowing unfree and nvidia acceptLicense to true

Edit: SOLVED

[ksvivek]

I tried the first config with nvidia 390, system boots but stays in lock screen. Tried entering both wayland and x11 session, it just brings me back to the loxl screen .
I’ll try with nvidia 470 now

[ksvivek]

Update:
Even with 470 this didn’t work, although from online search found that 390 is for my GPU.

And as someone might have said, the modern linux kernels might not work well with my old GPU . I wish there’s a working solution for my GPU driver for modern kernels

[TLATER]

@ksvivek

I tried the first config with nvidia 390, system boots but stays in lock screen. Tried entering both wayland and x11 session, it just brings me back to the loxl screen .

That’s… Not bad news tbh. You should be getting some kind of logging, if you switch to a tty with Ctrl+Alt+F5, log in there and get the latest logs for your user with journalctl --user --boot you might get some hints. Or even journalctl --boot, depending on when exactly this fails.

@ksvivek

I wish there’s a working solution for my GPU driver for modern kernels

The solution is to buy GPUs from companies that don’t lock you into proprietary drivers that they then discontinue a few years down the line But yeah, you’d have to have known 10 years ago. Vote for people who might force right-to-repair style legislation?

You can of course use nouveau, but I don’t know to what extent nouveau supports multi-gpu setups, or even how to begin configuring that to render videos for you, let alone on a 10 year old GPU from an era when nouveau was notoriously awful.

[ksvivek]

@TLATER

That’s… Not bad news tbh. You should be getting some kind of logging, if you switch to a tty with Ctrl+Alt+F5, log in there and get the latest logs for your user with journalctl --user --boot you might get some hints. Or even journalctl --boot, depending on when exactly this fails.

Ok I’ll try this.

@TLATER

The solution is to buy GPUs from companies that don’t lock you into proprietary drivers that they then discontinue a few years down the line But yeah, you’d have to have known 10 years ago. Vote for people who might force right-to-repair style legislation?

Just curious:
What about the current nvidia driver scenario? Is their open source drivers coming up well?
Also if I buy new PC / lap in future which graphics are well supported in Linux?

[TLATER]

@ksvivek

What about the current nvidia driver scenario? Is their open source drivers coming up well?

Hard to say in practice. The driver is still out-of-tree, so it won’t be maintained alongside the kernel. It’s very possible it will end up bitrotting if nvidia abandon it.

The userspace components are also still fully proprietary, so…

OTOH they have murmurred that they would like to upstream the driver eventually. Not that that has materialized.

There’s also renewed interest in a third party driver, which is going well as far as I’ve heard. Having an open kernel module to base that on likely helps a lot. But even with that you’re still looking at significant uncertainty in terms of the longer term future, given that it’s not actually working yet.

Overall the fact that a lot of functionality has moved into the GSP should make things a bit less problematic in the future, though?

@ksvivek

Also if I buy new PC / lap in future which graphics are well supported in Linux?

Both intel and AMD provide in-tree mesa drivers. For GPUs a consumer would buy it’s only nvidia who play silly games.