DEV Community

Arseny Zinchenko
Arseny Zinchenko

Posted on • Originally published at rtfm.co.ua on

Arch Linux: NVIDIA driver installation for GeForce GTX 1050 Ti

#archlinux #nvidia #linux #drivers

The main goal when I was buying a new PC with NVIDIA GeForce GTX 1050 Ti was an ability to play games on Arch Linux without the necessity to reboot into Windows.

For this need to install NVIDIA drivers so Linux will use it instead of nouveau (although I didn’t even try to run any game with nouveau).

After a couple of my attempts to install NVIDIA drivers on my laptop a few years ago, I expected issues here as well and was ready for long and deep dive into X.Org configs and logs. Actually, I started writing this post even before I faced with these installation issues.

Suddenly, everything went almost without problems – the only one I faced with was caused by my own mistake, see about it below.

Documentation is here>>>.

NVIDIA driver installation – attempt #1 (wrong)

Find a video card and a driver used now::

[setevoy@setevoy-arch-pc ~]$ lspci -k | grep -A 2 -E "(VGA|3D)"
01:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] (rev a1)
Subsystem: ASUSTeK Computer Inc. GP107 [GeForce GTX 1050 Ti]
Kernel driver in use: nouveau
Enter fullscreen mode Exit fullscreen mode

Check old cards list here here>>>.

Actually, the only issue was caused by the fact that I was tempted by the “lts” postfix in the nvidia-lts package name.

Install it ( no ):

[setevoy@setevoy-arch-pc .config]$ sudo pacman -S nvidia-lts
Enter fullscreen mode Exit fullscreen mode

Reboot PC and absolutely expected for me as I was waiting for problems – Openbox won’t start.

Check the driver used now:

[setevoy@setevoy-arch-pc ~]$ lspci -k | grep -A 2 -E "(VGA|3D)"
01:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] (rev a1)
Subsystem: ASUSTeK Computer Inc. GP107 [GeForce GTX 1050 Ti]
Kernel modules: nouveau
Enter fullscreen mode Exit fullscreen mode

Kernel modules: nouveau

Ummm…

And X.Org log /var/log/Xorg.0.log:

...
[    32.079] (==) Matched nouveau as autoconfigured driver 0
[    32.079] (==) Matched nv as autoconfigured driver 1
[    32.079] (==) Matched modesetting as autoconfigured driver 2
[    32.079] (==) Matched fbdev as autoconfigured driver 3
[    32.079] (==) Matched vesa as autoconfigured driver 4
[    32.079] (==) Assigned the driver to the xf86ConfigLayout
[    32.079] (II) LoadModule: "nouveau"
[    32.107] (WW) Warning, couldn't open module nouveau
[    32.107] (EE) Failed to load module "nouveau" (module does not exist, 0)
[    32.107] (II) LoadModule: "nv"
[    32.107] (WW) Warning, couldn't open module nv
[    32.107] (EE) Failed to load module "nv" (module does not exist, 0)
[    32.107] (II) LoadModule: "modesetting"
[    32.107] (II) Loading /usr/lib/xorg/modules/drivers/modesetting\_drv.so
[    32.123] (II) Module modesetting: vendor="X.Org Foundation"
[    32.123]    compiled for 1.20.3, module version = 1.20.3
[    32.123]    Module class: X.Org Video Driver
[    32.123]    ABI class: X.Org Video Driver, version 24.0
[    32.123] (II) LoadModule: "fbdev"
[    32.123] (WW) Warning, couldn't open module fbdev
[    32.123] (EE) Failed to load module "fbdev" (module does not exist, 0)
[    32.123] (II) LoadModule: "vesa"
[    32.123] (WW) Warning, couldn't open module vesa
[    32.123] (EE) Failed to load module "vesa" (module does not exist, 0)
[    32.123] (II) modesetting: Driver for Modesetting Kernel Drivers: kms
[    32.123] (EE) open /dev/dri/card0: No such file or directory
[    32.123] (WW) Falling back to old probe method for modesetting
[    32.123] (EE) open /dev/dri/card0: No such file or directory
[    32.123] (EE) Screen 0 deleted because of no matching config section.
[    32.123] (II) UnloadModule: "modesetting"
[    32.123] (EE) Device(s) detected, but none match those in the config file.
[    32.123] (EE)
Fatal server error:
[    32.123] (EE) no screens found(EE)
...
Enter fullscreen mode Exit fullscreen mode

Okay… Google? :-)

Found this>>> discussion – let’s try.

Need to update initramfs as mentioned here>>> and here>>>.

Edit the /etc/mkinitcpio.conf, add modules there:

...
MODULES=(nouveau, nv)
...
Enter fullscreen mode Exit fullscreen mode

Build a new kernel with a custom name to not overwrite the existing one:

[root@setevoy-arch-pc setevoy]# mkinitcpio -g /boot/linux-nv.img
==> Starting build: 4.20.11-arch2-1-ARCH
-> Running build hook: [base]
-> Running build hook: [udev]
-> Running build hook: [autodetect]
-> Running build hook: [keyboard]
-> Running build hook: [keymap]
-> Running build hook: [modconf]
-> Running build hook: [block]
-> Running build hook: [lvm2]
-> Running build hook: [filesystems]
-> Running build hook: [fsck]
==> ERROR: module not found: `nouveau,'
==> ERROR: module not found: `nv'
==> Generating module dependencies
==> Creating gzip-compressed initcpio image: /boot/linux-nv.img
==> WARNING: errors were encountered during the build. The image may not be complete.
Enter fullscreen mode Exit fullscreen mode

Well… There is obviously something wrong with this solution and as I did it in the evening – didn’t want to waste time looking for reasons and/or correct modules names.

NVIDIA driver installation – attempt #2 (correct)

Check NVIDIA-related packages installed now and pay attention to descriptions:

[root@setevoy-arch-pc setevoy]# pacman -Qs nvidia
local/libvdpau 1.1.1+3+ga21bf7a-1
Nvidia VDPAU library
local/nvidia-lts 1:418.43-1
NVIDIA drivers for linux-lts
local/nvidia-utils 418.43-2
NVIDIA drivers utilities
Enter fullscreen mode Exit fullscreen mode

NVIDIA drivers for linux-lts

And here I got my mistake: the kernel’s package linux-lts != linux and respectively nvidia-lts != nvidia and this is not a driver version with Long term Support - but just a driver for another kernel’s version.

Remove this driver:

[setevoy@setevoy-arch-pc ~]$ sudo pacman -Rsn nvidia-lts
Enter fullscreen mode Exit fullscreen mode

Install the common nvidia package:

[setevoy@setevoy-arch-pc ~]$ sudo pacman -S nvidia
Enter fullscreen mode Exit fullscreen mode

Reboot PC one more time and check the driver used now:

[setevoy@setevoy-arch-pc ~]$ lspci -k | grep -A 2 -E "(VGA|3D)"
01:00.0 VGA compatible controller: NVIDIA Corporation GP107 [GeForce GTX 1050 Ti] (rev a1)
Subsystem: ASUSTeK Computer Inc. GP107 [GeForce GTX 1050 Ti]
Kernel driver in use: nvidia
Enter fullscreen mode Exit fullscreen mode

Kernel driver in use: nvidia

Okay.

Monitors configuration

The last thing to do now is to update my monitors’ settings.

With the nouveau they had names:

[setevoy@setevoy-arch-pc ~]$ cat .config/openbox/autostart
xrandr --output HDMI-1 --primary
xrandr --output HDMI-1 --left-of DVI-D-1
Enter fullscreen mode Exit fullscreen mode

But now:

[setevoy@setevoy-arch-pc .config]$ xrandr --listmonitors
Monitors: 2
0: +DVI-D-0 1920/530x1080/300+0+0  DVI-D-0
1: +HDMI-0 1920/531x1080/299+1920+0  HDMI-0
Enter fullscreen mode Exit fullscreen mode

Update Openbox’s config and set them with the new names:

[setevoy@setevoy-arch-pc ~]$ cat .config/openbox/autostart
# nouveau
xrandr --output HDMI-1 --primary
xrandr --output HDMI-1 --left-of DVI-D-1

# nvidia
xrandr --output HDMI-0 --primary
xrandr --output HDMI-0 --left-of DVI-D-0


# tools

setxkbmap -layout us,ru -option grp:lctrl\_lshift\_toggle,grp\_led:scroll -variant winkeys &
volumeicon &
qxkb &
Enter fullscreen mode Exit fullscreen mode

In fact, better to do it with the udev rules – but still can’t find to realize it.

Done.

Similar posts

Top comments (5)

Collapse
 
mike239x profile image
Mike Lezhnin • • Edited

Hmmm, I am right now trying to use nvidia drivers on exactly the same video card, but even after deleting nouveau & mesa and installing nvidia I still get lspci saying the kernel module in use is nouveau :(
More googling it is.

P.S. ended up using nouveau for drivers, + installed nvidia + commented out that one place that blacklists nouveau. I still want to get cuda working, but having a working xorg seems more important right now :)

Collapse
 
cokesandtokes profile image
whoami •

Did you delete kms from mkinitcpio?

Collapse
 
fxos profile image
electric-otter •

You may create a basic udev rule, # Create a file in /etc/udev/rules.d/99-monitor-names.rules
SUBSYSTEM=="drm", KERNEL=="card0", ATTR{devname}=="HDMI-A-0", NAME="HDMI-0"
SUBSYSTEM=="drm", KERNEL=="card1", ATTR{devname}=="DVI-D-0", NAME="DVI-D-0"
And you can use these directories: /etc/udev/rules.d/99-monitor-names.rules, check your card number, card0 or card1, etc, and use KERNEL==cardnumber, open terminal (for example, Konsole), and type xrandr --listmonitors
should look like this Monitors:(number)
Connected device with graphics port for example: HDMI. or you could enter sudo nano /etc/udev/rules.d/99-monitor-names.rules
but if your a root user you can do nano /etc/udev/rules.d/99-monitor-names.rules
Always check your GPU configs. as they could be corrupted. To apply these changes type sudo udevadm control --reload, hopefully it fixes your arch linux issues! :)

Collapse
 
rduttshukla profile image
RIshi Dutt Shukla •

Can you clarify one thing, is this advice for notebook GTX 1050 Ti or desktop one?

Collapse
 
setevoy profile image
Arseny Zinchenko •

Hi, Rishi.
Sorry for the long reply - didn't notice notification.

Answering your question - I did this installation on my PC, not a laptop/notebook.

Read next

thenjdevopsguy profile image

Using Nvidia GPUs With Docker In 5 Minutes

Michael Levan -

jeyasurya2201 profile image

some key default port numbers that a DevOps engineer should be familiar with

jeya surya -

sunnybhambhani profile image

Quick guide to install minikube and enable ingress controller on Ubuntu

Sunny Bhambhani -

ayush7614 profile image

How to deploy Llama-3.1-Nemotron-70B-Instruct on a Virtual Machine in the Cloud?

Ayush kumar -