You are not logged in.
Been having this issue for a long time, including after reinstalling arch from scratch. Arch hangs at Loading Initial Ramdisk whenever i install nvidia, and is fine when i remove nvidia.
I've tried:
- blacklisting noveau
- adding nvidia, nvidia_modeset, nvidia_uvm, nvidia_drm to mkinitcpio.conf
- nvidia and nvidia-dkms
i've made sure to:
- get linux-headers first
- run mkinitcpio -P wherever necessary
journalctl boot log is available at http://termbin.com/mzwx. Nothing seems to be wrong in it, as far as I know...
Last edited by Phx_Blaster007 (2025-02-15 04:45:48)
Offline
Looks ok, what are you expecting to boot into? A TTY or a display manager and if so which? Because there are no traces of the latter but an attempt at starting a graphical target that doesn't go anywhere.
Which exact GPU? If you want to reach a TTY it could help to add
nvidia-drm.modeset=1 nvidia-drm.fbdev=0to your kernel parameters: https://wiki.archlinux.org/title/Kernel_parameters
Last edited by V1del (2025-02-13 21:46:14)
Offline
Looks ok, what are you expecting to boot into? A TTY or a display manager and if so which? Because there are no traces of the latter but an attempt at starting a graphical target that doesn't go anywhere.
Which exact GPU? If you want to reach a TTY it could help to add
nvidia-drm.modeset=1 nvidia-drm,fbdev=0to your kernel parameters: https://wiki.archlinux.org/title/Kernel_parameters
I've just been trying to boot into a TTY for now, I keep having to use my live installation to fix it.
That being said, progress! It boots when I set fbdev=0. I need it to be fbdev=1 though, because I do plan to have a window manager.
GPU is a GTX 1070.
Last edited by Phx_Blaster007 (2025-02-13 20:40:08)
Offline
I need it to be fbdev=1 though, because I do plan to have a window manager.
No, doesn't.
The combination of those two kernel parameters will allow you to run a full graphical stack.
Older nvidia GPUs tend to struggle w/ fbdev=1 (the current default) a lot, so you'll likely run it this way since it's not clear whether nvidia will ever get this straight.
Fwwi, the correct parameters are
nvidia_drm.modeset=1 nvidia_drm.fbdev=0(dash/underscore are irrelevant, comma/period is not)
Offline
Whoops, fixed that typo.
I actually wonder why the edit in the wiki was made, that suggests fbdev is in any way relevant to running any graphical setup... probably a generic driver bug getting mislabeled -- edit: ah read the linked reference yes that was to fix the 6.11 fb crazyness, afaik this isn't strictly necessary anymore.
Last edited by V1del (2025-02-13 21:50:04)
Offline
Here's a flamewar: https://wiki.archlinux.org/title/Talk:N … al_support
fbdev will be required to take over from the simplydumb device.
no simplydumb device means no requirement to take anything over and the modeset=1 hack to suppress that has been restored; so the wiki entry seems beyond misleading.
From all I've ever seen:
fbdev does not properly work on older generations and disabling it and suppressing the simplydumb device will be necessary and it does not seem to have changed w/ the 570xx drivers either.
Offline
I really want to believe that simpledrm isn't running and that I don't need fbdev=1, but it doesn't seem to be the case.
Hyprland is extremely laggy, and cpu usage spikes anytime I do anything. I've done everything on https://wiki.hyprland.org/Nvidia/ and made sure my nvidia modules and kernel settings are loaded properly. I'm running via uwsm, but I've tried running normally as well (and either way it shouldn't matter).
People in the hyprland discord seem to suggest that this lag is a typical issue and is usually fixed by fbdev=1, but it shouldn't make sense in this context...
hyprland log: https://termbin.com/9v4w
hyprland conf: https://termbin.com/idqd
/etc/mkinitcpio.conf: https://termbin.com/ssjc
uwsm env conf: https://termbin.com/vad1
uwsm env-hyprland conf:
export HYPRCURSOR_SIZE=24/etc/modprobe.d/nvidia.conf:
options nvidia_drm modeset=1 fbdev=0/etc/modprobe.d/disable-nouveau.conf:
blacklist nouveau
options nouveau modeset=0Last edited by Phx_Blaster007 (2025-02-14 18:28:06)
Offline
Luckily you don't have to believe, you can just query the system.
glxinfo -B
eglinfo -B
vulkaninfo --summaryBut your ovrhypdland log says
[LOG] Renderer: Nvidia version file exists
[LOG] Renderer: Read nvidia version 570.86.16
[LOG] Renderer: Parsed nvidia major version: 570so this does not seem to be the problem.
export __GL_GSYNC_ALLOWED=1Do you have an output capable of this and have your tried to disable it?
How do westong or sway or so perform for a contrast?
Offline
glxinfo.txt: https://termbin.com/btxg
vulkaninfo.txt: https://termbin.com/qlzj
Interestingly enough, eglinfo consistently prints some errors and then segfaults:
kmsro: driver missing
libEGL warning: egl: failed to create dri2 screen
kmsro: driver missing
libEGL warning: egl: failed to create dri2 screen
libEGL warning: egl: failed to create dri2 screen
libEGL warning: egl: failed to create dri2 screen
[1] 4578 segmentation fault (core dumped) eglinfo -B > eglinfo.txtRegardless, here's eglinfo.txt: https://termbin.com/qlzj (cutoff because of the seg fault)
I do have a GSync monitor, but I have tried disabling it and still have the same issue.
Sway doesn't support standard nvidia drivers (only nouveau), but weston, gnome wayland, and cosmic desktop all run great.
Offline
Solved! Fixed by setting fbdev=1 and modeset=1 in grub instead of via modprobe.
Last edited by Phx_Blaster007 (2025-02-15 04:45:19)
Offline