AMD Has An Important Suspend/Resume Fix With Linux 5.15

[gfunk]

On my old X1 carbon 5th gen, I never noticed any drain overnight, it would have been less than 5%

[Paradigm Shifter]

On my old X1 carbon 5th gen, I never noticed any drain overnight, it would have been less than 5%

Fair play. It doesn't appear uniform anyway. The laptop I'm referring to explicitly in the post above was an Acer, a "full fat" Core 2 Duo 11.6" job (thick as hell and with a horribly low screen resolution but had an amazing battery life for the time).

I'll put my LG Gram to sleep tonight, rather than shutting it down and see how much if any drain there is, because now I'm curious.

[Veerappan]

I find this article really timely. I hadn't checked my RSS feeds this afternoon yet, and spent an hour or so tonight checking to see if I could diagnose why my power LED, camera/microphone shutoff LEDs, and my brightness controls weren't working after suspend. I wasn't sure if there was a keyboard mapping issue post-resume, or if there was an ACPI issue with the way that resume was being handled.

Good to know that I can stop looking, and can just rebuild a kernel with a patch to some of the same code I already had open in an editor...

Is Linux capable of using s0ix for runtime powersaving on AMD hardware nowadays? I remember it resulted in significant improvements on Intel.

The 5.14 kernel is the first version that supported s0ix on the Renoir platform (e.g, my Ryzen 4700U-based HP Envy x360 13") out of the box. I've been watching the patches/activity around this significant interest for about a year now (since I got the laptop and reported the bug that ended up with this getting implemented/fixed).

Most of the activity around reporting/implementing this was done here: HP Envy x360 w/ 4700U failing to suspend (#1230) · Issues · drm / amd · GitLab

[khnazile]

Good, maybe they will switch Vega power management crashes in the future too ?

I don't think so, they care only about shiny new products. If you're a dedicated vega user, switch to rdna or polaris, dependin on what you can afford. If you're integrated/laptop user... Well, you will have to suffer until rdna2 APUs become available

[Lech]

I don't think so, they care only about shiny new products. If you're a dedicated vega user, switch to rdna or polaris, dependin on what you can afford. If you're integrated/laptop user... Well, you will have to suffer until rdna2 APUs become available

Well, problem is that Vega 56 should be powerful enough for my needs and if i can run cyberpunk on ultra in ~50 fps on 1080p, why should i change the gpu ? I'll sooner or later upgrade to some raytracing card, but current offer is not good enough in my book, and it will be acceptable in a generation or two.
The gpu hang is annoying, but that's it. What is worse, it works perfectly fine on windows, so it's not a hardware problem.

[timrichardson]

On my old X1 carbon 5th gen, I never noticed any drain overnight, it would have been less than 5%

I have a Carbon X1 gen 9, intel obviously. Modern suspend works really well, I never thought it possible. Lenovo has confidently labelled it the Windows/Linux option in the UEFI settings. S3 is there too ("Linux"). I hope these AMD fixes benefit the AMD IdeaPad my son has. We have it patched to run S3 and it works, but we have to use an older kernel.

[khnazile]

why should i change the gpu ?
The gpu hang is annoying, but that's it.

I don't know how do you live with it. Those hangs drive me mad so badly, that eventually I lost my motivation to try running games at all. What's the point, if it can crash any minute, and instead of fun you get frustration? When I realized this problem, I ordered RX550, and had no problems ever since. Yes, it's much slower, but my user experience with it still was much better than with Vega.

[Schugy]

My HP 15s-eg0355ng drains 15% even when shut down/halted. On top they closed my request for the UEFI bin file with a friendly "not supported". All UEFI upgrades are buried in exe files. That's the downside of a 300 € laptop.

[electropura]

The 5.14 kernel is the first version that supported s0ix on the Renoir platform (e.g, my Ryzen 4700U-based HP Envy x360 13") out of the box. I've been watching the patches/activity around this significant interest for about a year now (since I got the laptop and reported the bug that ended up with this getting implemented/fixed).

That's interesting. S2idle works flawlessly with the 5.13 kernel on my Lenovo Yoga slim 7 with a Ryzen 5800U. Unfortunately something broke suspend in 5.14 so I had to go back to a 5.13 kernel again. What is the best way to report this regression? The distro or try to report directly upstream?

[Veerappan]

That's interesting. S2idle works flawlessly with the 5.13 kernel on my Lenovo Yoga slim 7 with a Ryzen 5800U. Unfortunately something broke suspend in 5.14 so I had to go back to a 5.13 kernel again. What is the best way to report this regression? The distro or try to report directly upstream?

Hmm, I'm not sure where the best place to report is. My initial issue appeared to be due to a stacktrace in the amdgpu drm module, so I reported it there. It ended up being completely unrelated to the issue at hand, but it got the right people involved.

What sort of behavior are you experiencing? It worked in 5.13 for you, but 5.14 is broken. Do you have the ability to try any of the RC kernels, or bisect the regression? If I/we know more about what is no longer working, we might be able to point you in the right direction.