This is a simple readme of how I was able to configure the nvidia drivers so a nvidia gpu is suspended except when it is needed by an application, such as cuda.
I only tested this on an Asus Zephyrus GU502gv using ubuntu 19.10 using the nvidia drivers 435.21
This assumes that the nvidia drivers are already installed. (check this by running the command nvidia-smi.
If you have an output in the terminal then you are good to procede.
- Configure nvidia-prime to on-demand mode
- Configure Xorg to use the intel integrated graphics
- Configure power management from the nvidia drivers
First you need to set the gpu to be on-demand.
Run these commands
$ sudo prime-select on-demand
$ sudo reboot
First check to see if your computer is using the nvidia graphics or not by running
$ nvidia-smi
you should get an output similar to below
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 435.21 Driver Version: 435.21 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 2060 Off | 00000000:01:00.0 Off | N/A |
| N/A 37C P0 N/A / N/A | 0MiB / 5934MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
under Processes it should say No running processes found.
If you do have xorg running (like I did) then you need to configure xorg explicitely
edit the file /etc/X11/xorg.conf to have the following contents
Section "ServerLayout"
Identifier "layout"
Screen 0 "iGPU"
Option "AllowNVIDIAGPUScreens"
EndSection
Section "Device"
Identifier "iGPU"
Driver "modesetting"
EndSection
Section "Screen"
Identifier "iGPU"
Device "iGPU"
EndSection
#Section "Device"
# Identifier "dGPU"
# Driver "nvidia"
#EndSection
source: http://download.nvidia.com/XFree86/Linux-x86_64/440.31/README/primerenderoffload.html
Then you will need to restart your computer. or restart gdm or lightdm sudo systemctl restart gdm
Make sure you know what you are doing when you edit xorg.conf files. It can mess with your computer. So be sure to make a backup of your config files before changing anything. I had to secure boot a few times and revert configuration files back to the previous state before I found the configuration nvidia suggests.
nvidia power management lets the driver suspend and wake the GPU as needed, but there are some hardware requirements
- This feature is supported only on notebooks.
- This feature requires system hardware as well as ACPI support (ACPI "_PR0" and "_PR3" methods are needed to control PCIe power). The necessary hardware and ACPI support was first added in Intel Coffeelake chipset series. Hence, this feature is supported from Intel Coffeelake chipset series.
- This feature requires a Turing or newer GPU.
- This feature is supported with Linux kernel versions 4.18 and newer. With older kernel versions, it may not work as intended.
- This feature is supported when Linux kernel defines CONFIG_PM (CONFIG_PM=y). Typically, if the system supports S3 (suspend-to-RAM), then CONFIG_PM would be defined.
Source: http://download.nvidia.com/XFree86/Linux-x86_64/440.31/README/dynamicpowermanagement.html
If your computer fits the hardware requirements (they may change in the future, so check the docs if you are unsure)
you can edit the driver to change power management.
From the nvidia docs
Option "NVreg_DynamicPowerManagement=0x00"
With this setting, the NVIDIA driver will only use the GPU's built-in power management so it
always is powered on and functional. This is the default option, since this feature is a new
and highly experimental feature. Actual power usage will vary with the GPU's workload.
Option "NVreg_DynamicPowerManagement=0x01"
With this setting, the NVIDIA GPU driver will allow the GPU to go into its lowest power state
when no applications are running that use the nvidia driver stack. Whenever an application
requiring NVIDIA GPU access is started, the GPU is put into an active state. When the
application exits, the GPU is put into a low power state.
Option "NVreg_DynamicPowerManagement=0x02"
With this setting, the NVIDIA GPU driver will allow the GPU to go into its lowest power state
when no applications are running that use the nvidia driver stack. Whenever an application
requiring NVIDIA GPU access is started, the GPU is put into an active state. When the
application exits, the GPU is put into a low power state.
Additionally, the NVIDIA driver will actively monitor GPU usage while applications using the
GPU are running. When the applications have not used the GPU for a short period, the driver
will allow the GPU to be powered down. As soon as the application starts using the GPU, the
GPU is reactivated.
It is important to note that the NVIDIA GPU will remain in an active state if it is driving a
display. In this case, the NVIDIA GPU will go to a low power state only when the X
configuration option HardDPMS is enabled and the display is turned off by some means - either
automatically due to an OS setting or manually using commands like xset.
Similarly, the NVIDIA GPU will remain in an active state if a CUDA application is running.
To enable Dynamic Power Management edit the file /etc/modprobe.d/nvidia.conf
and add the setting you want.
My configuration is
modprobe nvidia "NVreg_DynamicPowerManagement=0x02"