nvidia-powerd.service needs a reboot to change the power limit state on AMD lenovo gaming laptops

[ghost]

NVIDIA Open GPU Kernel Modules Version

530.41.03

Does this happen with the proprietary driver (of the same version) as well?

Yes

Operating System and Version

Fedora release 37 (Thirty Seven)

Kernel Release

Linux DESKTOP-MPKTHBJ 6.2.10-200.fc37.x86_64 #1 SMP PREEMPT_DYNAMIC Thu Apr 6 23:30:41 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

Hardware: GPU

GPU 0: NVIDIA GeForce RTX 3050 Laptop GPU (UUID: GPU-b2f720a1-2b4e-9b17-4383-76f9361248a2)

Describe the bug

nvidia-powerd.service needs a full system reboot to change the power limit of the GPU from 64-66W in the laptop's balanced mode to 85W that the laptops' performance mode should use

If i press fn+Q and change to performance mode without restarting the laptop will still use 64W
After a system restart the power limit changes to 85W but if i change back to balenced/power saver mode the power limit stays at 85W which is very bad for the GPU temp and may cause hardware damage

The problem here is have to reboot each time you need to change the power limit while on Windows it changes dynamically just by pressing FN + Q

To Reproduce

Change your lenovo legion/ideapad laptop to performance mode from balenced without a system retsrat and check the GPU power draw

Bug Incidence

Always

nvidia-bug-report.log.gz

nvidia-bug-report.log.gz

More Info

Laptop specs:

Lenovo Ideapad Gaming 3 15ACH6
AMD Ryzen 5 5600H
RTX 3050 Mobile 85W
16 GB RAM & 1 TB SSD M.2

[beh-10257]

I would guess your problem is the same or similar to my problem
#491
can you check that like if the driver even realize the powersource change
most likely internally it's the same bug lool

[ghost]

I would guess your problem is the same or similar to my problem #491 can you check that like if the driver even realize the powersource change most likely internally it's the same bug lool

it says 0 even when the laptop draws 85W

[beh-10257]

@kleidiss
btw 0 is plugged 1 is unplugged
just watch "nvidia-settings -q GPUPowerSource"
and plug your laptop I guess give it 10 seconds to make sure
unplug give it another 10 seconds if there is no switch in the open driver you basically have the same problem as me

[ghost]

@kleidiss btw 0 is plugged 1 is unplugged just watch "nvidia-settings -q GPUPowerSource" and plug your laptop I guess give it 10 seconds to make sure unplug give it another 10 seconds if there is no switch in the open driver you basically have the same problem as me

Ok so i unplugged it went to 1 and plugged back in it went to 0 again

[beh-10257]

@kleidiss you used xorg for that test right

• yeah there is a difference cause the test I told you to do actually works under wayland for some reason
• if you are using gnome or kde I guess you can switch in the login manager

[beh-10257]

[update]
since you game me the idea to just the power usage without unplugging anything fps in minecraft will be 120 with shaders and power draw 50 and the mhz 2000
after unplugging fps to 10 and power draw to 15w ( why is that the case for gods sake just give to the gpu 30w at least )
and the mhz to 300 and sometimes 340
and the mhz and power draw will stay exactly the same even after replugging the laptop

[ghost]

* yeah there is a difference cause the test I told you to do actually works under wayland for some reason

Oh i use wayland on fedora
Yeah switching to Xorg i see now but i dont think is related to my issue because i also tested the proprietary drivers and there was no such issue there even on Xorg but my specific issue also happens on proprietary drivers

[ghost]

[update] since you game me the idea to just the power usage without unplugging anything fps in minecraft will be 120 with shaders and power draw 50 and the mhz 2000 after unplugging fps to 10 and power draw to 15w ( why is that the case for gods sake just give to the gpu 30w at least ) and the mhz to 300 and sometimes 340 and the mhz and power draw will stay exactly the same even after replugging the laptop

15w lmao , might as well turn off the gpu at that point

[beh-10257]

• 15w lmao , might as well turn off the gpu at that point
  exactly

Yeah switching to Xorg i see now

at least the problem is not just for me
so yeah I guess your response confirms #491

my specific issue also happens on proprietary drivers

two issue related to power in two days that's hilarious

[ghost]

two issue related to power in two days that's hilarious

Not to mention that on the Unigine Heaven OGL benchmark i get 80-100 points less on Linux while using the same power draw and clocks as Windows (I checked with mangohud and geforce overlay thing on windows) These crappy drivers are also burning power for no reason

I notice a fps loss on DXVK as well while using these drivers on Linux but thats more acceptable than a native benchmark

[ghost]

Correction: You just need to reload the service for it to update the power and clock speed limit

Still annoying having to reload it each time tho

[beh-10257]

@kleidiss
Wait there is a service for that
What is it
Probably that would solve my problem as well

[ghost]

@kleidiss Wait there is a service for that What is it Probably that would solve my problem as well

nvidia-powerd.service
It manages dynamic boost meaning it takes power from the CPU when the CPU isn't using it and gives it to the GPU
I don't think that is related to battery tho

[ghost]

Just reload the service when unplugged and see if the power draw changes
If it does then it's confirmed we have basically the same problem

systemctl stop nvidia-powerd.service and than just start it again

[beh-10257]

well it seems that my system doesn't support that in the first place so yeah we don't have the same problem
I guess good for you for your temporary solution

[saltyming]

Restarting nvidia-powerd service without rebooting causes dgpu unable to go to d3cold state, making power drain even when it's not being used. So it's still not a ideal solution. My Lenovo laptop is 15ARH7, Ryzen 6600H with NVIDIA RTX3050.

EDIT: Need more testing.

[NiiSV811]

Hi there. I have the same Laptop and I have problems with suspend.
Steps to reproduce the problem:

• Plug the laptop
• suspend
• Unplug laptop while it's suspend
• Resume the laptop, now turn off, and it won't turn off.
  Also if you plug HDMI and suspend laptop. It won't suspend. It's weird.

P.D: Sorry for my English. I'm learning. And Thank you if you could help me, I am a beginner in Linux world

[ghost]

P.D: Sorry for my English. I'm learning. And Thank you if you could help me, I am a beginner in Linux world

I usually have my laptop plugged in so idk

Make a new issue for this too so more people see it

[beh-10257]

have anything changed in your part for the past 2 weeks
for me even though I've switched to open beta drivers no difference

[ghost]

New drivers same bug

[beh-10257]

lol I will wait for these drivers to drop in the aur in my side and update my problem as well
this repo is pathetic
at least tell us that you have putted the bug in your internal bug tracker

[onenowy]

I have a similar problem. On hp vicuts 16-d1xxx laptop with patch (https://lore.kernel.org/platform-driver-x86/[email protected]/), It needs restarting nvidia-powerd.service after changing platform profile to change tgp.

[ghost]

@onenowy
Seems like this issue thread isn't being seen by the NV team
Better go explain this bug on this issue instead

#392

And yes we have the same bug
Changing platform profile doesn't do shit unless a restart

[amrit1711]

Dear All,
We have already filed a bug 4142071 for this issue and it has been also root caused.
We will integrate the fix in future release drivers and shall update accordingly.

[amrit1711]

Dear All,
Fix is available in latest release driver, please verify and share test results.

[onenowy]

It's working with 3060 mobile on hp victus 16 laptop.

[xAlpharax]

Dear All,
Fix is available in latest release driver, please verify and share test results.

Great! Works for me (Acer AN515-45) with a 3070.

[ghost]

The fix is on the driver from 3 weeks ago right?

[ghost]

Still not fixed on 535.113.01 for me
Switching to performance power profile and GPU is still stuck at 64 watts and not going up to 85 watts like its supposed to

This was tested on closed source driver so maybe the fix is not merged into that yet?

[amrit1711]

@kleidiss
Could you please help to share nvidia bug report from repro state.
Thanks in advance.

[ghost]

@amrit1711 Sorry for the late reply

using nvidia-open-dkms from arch repos 545.29.06 bug isnt fixed
nvidia-bug-report.log.gz

[ghost]

nvidia-smi -i 0 -q -d POWER

==============NVSMI LOG==============

Timestamp : Thu Dec 28 18:52:49 2023
Driver Version : 545.29.06
CUDA Version : 12.3

Attached GPUs : 1
GPU 00000000:01:00.0
GPU Power Readings
Power Draw : 64.93 W
Current Power Limit : 65.00 W
Requested Power Limit : N/A
Default Power Limit : 60.00 W
Min Power Limit : 1.00 W
Max Power Limit : 85.00 W
Power Samples
Duration : Not Found
Number of Samples : Not Found
Max : Not Found
Min : Not Found
Avg : Not Found
GPU Memory Power Readings
Power Draw : N/A
Module Power Readings
Power Draw : N/A
Current Power Limit : N/A
Requested Power Limit : N/A
Default Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A

[mguinhos]

I think i'm with a similar problem on my Dell G15 (5530 RTX 3050 6GB)

[luckyluca]

I'm having a similar problem with the Nvidia driver on Windows 11 on my Lenovo Legion 7 Slim AMD Nvidia 4060 laptop.
Is this related?
using nvidia-smi does return the following error:
Changing power management limit is not supported in current scope for GPU: 00000000:01:00.0.

https://www.nvidia.com/en-us/geforce/forums/geforce-laptops/6/541317/changing-power-management-limit-is-not-working-on-/

[amrit1711]

Hi All,
We would like to know if there are any other Legion 5 models apart from Lenovo Ideapad Gaming 3 15ACH6 which still exhibits this issue. If yes, please share model information.

[Gert-dev]

In case it helps: I have a Lenovo Legion Pro 7 16ARX8H since fairly recently and I can confirm this does not happen for me; I can see the power limit changing from 150 to 175 and back dynamically when demanding tasks run. This also still works when switching between different modes using Fn+Q and switching between the performance and quiet profiles dynamically switches from 80 Watt to 150 Watt in nvidia-smi.