Nvidia Optimus on Linux

If you’re using any distribution of Linux on your laptop with Nvidia GPU you know that Nvidia drivers are problematic. Why? There are several reasons:

1. Nvidia drivers are proprietary, which means their code is closed, and only Nvidia can release updates. So the community cannot do anything except wait until Nvidia decides to implement or fix something. There are open-source drivers, however, their performance is significantly worse than the proprietary ones.
2. Until recently there was no support for Nvidia Optimus which allows using your integrated video-card for casual applications and GPU for games and other GPU-heavy processes. This problem is semi-solved with PRIME which is the best case requires to log out and log in back.
3. Nvidia GPU consumes a lot of power when it’s not used. So there is no way to turn it off when it’s not used.

But now, the problem #2 and #3 is solved in drivers 435.xx by adding support for PRIME Render Offloading which allows rendering GPU heavy applications on GPU and others on the integrated. Also, the drivers introduced support for D3 Power Managment which makes GPU not consume power when it’s not used.

I tried to install those new drivers and see whether it’s working, and it DID WORK!

Unfortunately, there are few big and important requirements to make D3 power management to work:

1. Turing-generation GPU. Those are GPUs starting from GTX 1650, see the link for more details.
2. Intel 8th generation Coffee Lake processor

If your laptop (or PC) meets the requirements you can enable D3 power management mode, otherwise, you only can enable Nvidia Offloading which without D3 is not that useful on laptops as Nvidia GPUs consumes a lot of power and drain the battery very fast.

I tried to do this on my Dell XPS 15 7590 with GeForce GTX 1650 and Intel(R) Core(TM) i7-9750H. I’m running Manjaro Linux as my distro so the instruction bellow will use Manjaro specific tools.. For Ubuntu/Debian you probably can use apt with appropriate package names.

Let’s start!

1. Follow Nvidia driver’s instructions and remove built-in drivers and install driver 435xx or newer. In Manjaro you can do:

   sudo mhwd -i pci video-nvidia-435xx

2. After installation, reboot or logout to apply new changes, whatever works for you. If you correctly installed the drivers you should be able to see output for nvidia-smi command. Which looks like something like this:

   $ nvidia-smi
   Sat Nov 23 14:47:49 2019
   +-----------------------------------------------------------------------------+
   | NVIDIA-SMI 440.31       Driver Version: 440.31       CUDA Version: 10.2     |
   |-------------------------------+----------------------+----------------------+
   | 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 GTX 1650    Off  | 00000000:01:00.0 Off |                  N/A |
   | N/A   44C    P8    14W /  N/A |     38MiB /  3914MiB |      0%      Default |
   +-------------------------------+----------------------+----------------------+
   
   +-----------------------------------------------------------------------------+
   | Processes:                                                       GPU Memory |
   |  GPU       PID   Type   Process name                             Usage      |
   |=============================================================================|
   |    0     17281      G   /usr/lib/Xorg                                 37MiB |
   +-----------------------------------------------------------------------------+
       ```

3. The next step is to enable GPU offloading. You can do it by performing these official instructions from Nvidia. For Majaor, you can install optimus-manager and set hybrid mode (ignore the warning for now).

   sudo pacman -S optimus-manager
   optimus-manager --switch hybrid

4. (Optinal) Currently, no stable version of X.org or Wayland supports GPU offloading. However, for X.org, if you want it to run on GPU you can install the patched by Nvidia engineer X.org. For Manjaro just clone the repo, build, and install the package. After the installation log out and log in back to restart X.org server:

   cd "$(mktemp  -d)"
   git clone https://gitlab.freedesktop.org/aplattner/arch-xorg-servercd arch-xorg-server
   makepkg -s
   sudo pacman -U *.xz

5. Enable D3 Power Mode so your Nvidia GPU is turned off when not used by following guide from Nvidia (use Automated Setup). Reboot to apply the changes.

If you everything did correct, you should be able to offload processes to GPU by setting the following environment variables:

• __NV_PRIME_RENDER_OFFLOAD=1 for Vulkan apps
• __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia for OpenGL apps.

I created an prime alias and saved in .bashrc for settings those variables. To do it add this line to your .bashrc or .bash_profile or whatever shell config you’re using:

alias prime='__NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia'

And to verify it works try the following:

# mikael @ mikael-pc in /tmp/tmp.NNYn0gkreU [14:38:37]
$ prime glxinfo | grep -i "opengl renderer"OpenGL renderer string: GeForce GTX 1650/PCIe/SSE2
# mikael @ mikael-pc in /tmp/tmp.NNYn0gkreU [14:38:40]
$ glxinfo | grep -i "opengl renderer" OpenGL renderer string: Mesa DRI Intel(R) UHD Graphics 630 (Coffeelake 3x8 GT2)

As you can see, when I append prime the process uses Nvidia GPU and if I don’t it uses Intel GPU. To offload games in Steam to your GPU add those environment variables to your game launch options. If your game uses Vulkan API you can do: __NV_PRIME_RENDER_OFFLOAD=1 %command%

I hope this guide helped you to setup GPU Offloading and D3 power management state. If you have questions write in the comments, I’ll try to answer then.

Thanks for reading!