So, /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="quiet acpi_osi="!Windows 2017" acpi_osi="!Windows 2015" acpi_osi=Linux-Dell-Thunderbolt acpi_osi=Linux-Dell-Video"
Correct?
In grub, you’ll have to escape the quotes, like
acpi_osi="!Windows 2015"
done, then sudo update-grub, reboot, tested - nothing changed
D’oh. I’m out of ideas then. You can only bug dell support how this should work.
But are you sure that root of the problem is ACPI and not nvidia driver?
Thanks for all the help and suggestions anyway!
I appreciate your help
Sure? No. But since the P5200 is quite new and driver support has only been added with v384 and you tried drivers beginning with v390 (that’s why I asked) I find it rather unlikely for a driver regression that fast. Of course, you’ll never know. Be sure to send the nvidia-bug-report.log with a description of the problem to linux-bugs[at]nvidia.com, maybe they get more information out of it.
Yeah, hardware of this machine is really fresh…
Sent both to DELL & linux-bugs[at]nvidia.com
Hope they can help
Hello,
we currently have NVIDIA bug #2473619 opened by Dell describing similar symptoms on a Dell Precision 7730 system with Nvidia Quadro P5200 GPU. It sounds highly likely to be a duplicate of what you’re describing.
We’re actively looking into it, although right now there isn’t a fix available.
Thank you for letting me know, hopefully solution will come!
If you’ll need some help with testing i’m here for you
This particular effect when you switch to BAT and then back to AC doesn’t restore full GPU performance - seems to be solved for me with 5.0 Kernel + 418.24 drivers.
So if this wasn’t just happy accident - thanks for good work :)
Though performance on BAT graphics / cpu is still kinda low :(
Performance sticking at minimum clocks on battery is actually a feature, might be overridden by setting the desired PowerMizer settings.
Under hard load?
Not really.
Forcing it doesn’t change this behavior, if you’ve read the whole story and original thread (long and understandably hard to follow) which has led me here, i suspect it’s hardware specific issue maybe in combination with Manjaro.
Since on BAT under any kind of graphical load (even resizing window or something like that) my CPU for some reason just bumps down really hard.
It happens regardless of powermizer state / settings.
And also to be fair, even adaptive powermizer shouldn’t restrict highest modes.
Funniest part is that on Ubuntu it doesn’t happen at all…
Hi, Any news about NVIDIA bug #2473619?
Where I can check the status?
Thank you
Bug #2473619 should be fixed in 430.09
Where can we track bugs btw, it’s not public?
I’m on 430.26 and Kernel 5.1 now it works good, but as mentioned have been solved previously with 418.24 already (part of problem that AC → BAT → AC haven’t recovered from poor performance state).
Also there are no extreme CPU bumps down to 800mHz on graphical load anymore…
Which is definitely good, thank you very much for your work!
But performance on battery is still very poor (for example Firefox scroll is very laggy), which now looks even weirder since a lot of obvious things like CPU / GPU frequencies are effectively ruled out.
CPU is pretty much adequately reacts on heavy load, i can force it up to the limits and enable Intel Boost through TLP, so i don’t think CPU is problem here. Also it doesn’t throttle or something like that, according to journal.
GPU as far as nvidia-settings tells is not getting higher than level 2 PowerMizer & ~400 mHz grpahics clock (which i assume from experience shouldn’t look that laggy).
I can confirm that the bug has been fixed in one of the recent driver versions but now there seems to be a regression because it’s back again in 435.21 on Gentoo Linux.
I tested it with a Quadro P3200 in a Dell Precision 7530 (affected by this bug, too). Unplugging an AC Adapter causes the GPU to be throttled to ~ 200mhz, plugging the AC adapter back doesn’t change that.
acpid is running, nvidia-settings display the correct AC plugged state.
I tested it with ubuntu 18.04 and there the 435.21 worked as it should. Replaced the kernel source by a vanilla kernel on ubuntu to ensure that it’s not a custom kernel thing: still works. I have no idea what’s going wrong here.
Was anyone able to fix that issue outside ubuntu?