You are not logged in.
In some laptops like mine, this problem exists: Extremely_high_battery_usage_because_of_dedicated_GPU - ArchWiki
The wiki steps to resolve that (link above) look good (and are very clear), but as I was reading them, I wondered if this much simpler approach will work instead. Here's what I'm in the process of trying:
1. install nvidia-xrun
2. add the nvidia-xrun-pm.service (which it seems really should be included in the AUR package but isn't)
3. enable the nvidia-xrun-pm.service
Is there any reason why that can't accomplish the same result as all the steps written in the Wiki (link above)?
In more detail, here is what I did so far:
pacman -S nvidia-xrun
cat /etc/systemd/system/nvidia-xrun-pm.service
[Unit]
Description="Remove Nvidia GPU from kernel devices list and enable PM"
[Service]
Type=oneshot
Environment="TURN_OFF_GPU_ONLY=1"
ExecStart=/usr/bin/nvidia-xrun
[Install]
WantedBy=multi-user.target
systemctl enable nvidia-xrun-pm.serviceI built the AUR package myself which is why I installed it with "pacman -S". Also, I just ran a full system update so don't be concerned about the lack of "yu" args.
Offline
Should work, yes. Just try it and check your overall consumption with powertop.
Offline
Should work, yes. Just try it and check your overall consumption with powertop.
OK, I have to investigate a bit deeper:
# systemctl status nvidia-xrun-pm.service
● nvidia-xrun-pm.service - "Remove Nvidia GPU from kernel devices list and enable PM"
Loaded: loaded (/etc/systemd/system/nvidia-xrun-pm.service; enabled; vendor preset: disabled)
Active: failed (Result: exit-code) since Sun 2020-11-22 20:49:26 EST; 20min ago
Process: 722 ExecStart=/usr/bin/nvidia-xrun (code=exited, status=1/FAILURE)
Main PID: 722 (code=exited, status=1/FAILURE)
Nov 22 20:49:26 laghima systemd[1]: Starting "Remove Nvidia GPU from kernel devices list and enable PM"...
Nov 22 20:49:26 laghima nvidia-xrun[722]: This script must not be run as root
Nov 22 20:49:26 laghima systemd[1]: nvidia-xrun-pm.service: Main process exited, code=exited, status=1/FAILURE
Nov 22 20:49:26 laghima systemd[1]: nvidia-xrun-pm.service: Failed with result 'exit-code'.
Nov 22 20:49:26 laghima systemd[1]: Failed to start "Remove Nvidia GPU from kernel devices list and enable PM".From the github page Witko/nvidia-xrun: Utility to run separate X with discrete nvidia graphics with full performance I saw:
/etc/systemd/system/nvidia-xrun-pm.service systemd serviceBut apparently that won't work because of the above error "This script must not be run as root".
I will probably just go with the original steps at the Arch Wiki then.
Last edited by MountainX (2020-11-23 02:15:41)
Offline
Now I completed all the steps at the Wiki page (url in OP). I have this content in bbswitch.conf.
# cat /etc/modprobe.d/bbswitch.conf
options bbswitch load_state=0 unload_state=1I have enabled all the settings to disable the nvidia GPU at boot and I rebooted. The wiki says, "Check the power status of the GPU by running cat /proc/acpi/bbswitch."
# cat /proc/acpi/bbswitch
0000:3c:00.0 ONDoes this output mean the nvidia GPU is "ON"?
The output of that remains "ON" even after manually running:
tee /proc/acpi/bbswitch <<<OFFI have not worked with bumblebee in the past. Never had a laptop with a 2nd GPU before.
Last edited by MountainX (2020-11-23 03:12:19)
Offline
What's in your dmesg? there should be messages if bbswitch was successful in disabling the card
FWIW depending on how new the system is you might generally prefer foregoing all of this, installing the nvidia driver and enabling power saving as it's included in it: https://download.nvidia.com/XFree86/Lin … ement.html
Last edited by V1del (2020-11-23 08:11:38)
Offline
In my case I removed bbswitch, and just blacklisted the nvidia modules
/etc/modprobed.d/blacklist.conf
blacklist nvidia
blacklist nvidia_drm
blacklist nvidia_uvm
blacklist nvidia_modesetAnd created a service to put in auto the pci id
echo auto > /sys/bus/pci/devices/0000:01:00.0/power/controlThat should put the nvidia card in suspended mode :
$cat /sys/bus/pci/devices/0000:01:*/power/runtime_status
suspendedThe battery goes from ~3hrs to ~7hrs, and the temperature of that side of the laptop goes down .
The link that videl provides is useful when you are using the nvidia with prime render offload or reverse prime, and the nvidia modules are being used but still consuming energy depending of the card model from 1w (2xxx card models) to ~6-7w (1xxx card models).
The problem is that in my laptop I cannot wakeup again the card showing some pcihp error of card not found I have to reboot, but in my case that I use the nvidia card once a week to play or anything related to nvidia (ML) just commenting the blacklisted lines and rebooting is enough for me.
Offline