Arch Linux

barney

Member

Registered: 2020-12-14

Posts: 130

Problems with Nvidia + Hyprland

I have a laptop with Intel integrated graphics + Nvidia 1060 graphics card. I want to run Hyperland on Nvidia Graphics card. I have followed the official wiki and Arch Wiki about Wayland and Hyprland.
I have made two permanent links in /dev/dri for nvidia and intel cards, names nvidia-igpu and intel-igpu. I am using UWSM to start Hyprland. Contentc of the ~/.config/uwsm/env-hyprland is:

 #export WLR_DRM_DEVICES=/dev/dri/card0
export AQ_DRM_DEVICES="/dev/dri/nvidia-igpu:/dev/dri/intel-igpu"
#export __NV_PRIME_RENDER_OFFLOAD=1
export LIBVA_DRIVER_NAME=nvidia
export __GLX_VENDOR_LIBRARY_NAME=nvidia
#export __VK_LAYER_NV_optimus=NVIDIA_only
export NVD_BACKEND,direct
#export XDG_SESSION_TYPE=wayland
export SDL_VIDEODRIVER="wayland,x11"
#env = QT_QPA_PLATFORM,wayland
#env = GDK_BACKEND,wayland,x11
#env = MOZ_ENABLE_WAYLAND,1
#env = MOZ_DISABLE_RDD_SANDBOX,1
#env = MOZ_DBUS_REMOTE,1
export GBM_BACKEND=nvidia-drm
#env = GBM_BACKEND,nvidia
export HYPRCURSOR_THEME=rose-pine-hyprcursor
export HYPRCURSOR_SIZE=16
export HYPRSHOT_DIR="/home/barney/pictures/screenshots"

There are some variables that are commented, they are from some experiments that I have made reading REDDIT posts.
As mentioned on the official wiki page I have added nvidia.NVreg_PreserveVideoMemoryAllocations=1 to my kernel command line ( I am using UKIs).
I have this line:

MODULES=(nvidia nvidia_modeset nvidia_uvm nvidia_drm)

in my /etc/ mkinitcpio.conf file.
The problem is that resume from suspend is giving black screen. The power button is responding and I can turn off my laptop. Also not all applications are running  under Nvidia. If I start for example Thunar, firefox, spotify, Alacritty, Emacsclient I get this from nvidia-smi:

+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.126.09             Driver Version: 580.126.09     CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA GeForce GTX 1650 Ti     Off |   00000000:01:00.0 Off |                  N/A |
| N/A   40C    P8              1W /   50W |     309MiB /   4096MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|    0   N/A  N/A            1244      G   Hyprland                                 85MiB |
|    0   N/A  N/A            1280      G   /usr/bin/alacritty                       24MiB |
|    0   N/A  N/A            1281      G   /usr/lib/firefox/firefox                157MiB |
|    0   N/A  N/A            1289      G   Xwayland                                  2MiB |
+-----------------------------------------------------------------------------------------+

As you can see only firefox and Alacritty are running under Nvidia.
Here is my journal: http://0x0.st/PKnK.txt

V1del

Forum Moderator

Registered: 2012-10-16

Posts: 25,064

Re: Problems with Nvidia + Hyprland

Why do you want to run hyprland and normal desktop applications on the dedicated GPU? If you don't set any of these variables you should get the "proper" and intended variation of this setup that only demanding/gaming tasks use the GPU and normal desktop workflows can be kept to the integrated GPU that has much more direct access to the screen: https://wiki.archlinux.org/title/PRIME# … er_offload

You are potentially incuring a performance hit because the nvidia GPU still has to copy everything to your integrated GPU for display, unless you have a dedicated hdmi connection to an external screen.