Ok so I tried using mhwd to install 570xx like you described, it went smoothly, 570 is installed, however it did not fix my issue unfortunately. Is it possible to go back to 550xx, or would that be likely to break things?
It looks like only 570xx, 575xx (and greater) drivers will support the GeForce RTX 4070, so 550xx is out of the proverbial ball park.
I note that your BIOS has updates available.
While an update may not magically fix the current issue, itās nonetheless a consideration that might improve something related. The choice is yours.
The latest (non-Beta) BIOS for the Z370 AORUS Gaming 5 (rev. 1.0) appears to be F15. Search the Gigabyte site for instructions on how to apply it.
How can that be? I bought my 4070 about a week after it was released and the most current driver series was 515xx at the time. I was running 550xx until a few months ago when I started noticing this problem
Remember, these are the proprietary drivers from Nvidia, the only difference is that they are repackaged for Linux.
These are what I found on the Nvidia site that support the GeForce RTX 4070; there are no versions below 570.124.04 listed.
If what you say is accurate (and I have no reason to doubt it) then perhaps Iām wrong about that; Nvidia might have removed 550.xxx from their search results.
You havenāt responded regarding possibly updating the BIOS.
You claim the same issue arises with 580xx and now 570xx:
Here I see the 550xx mentioned and that a timeshift restore apparently āresolved the issueā. I suspect this may only be coincidence, however, you could arguably test that theory by following the same procedure used to install 570xx.
Howeverā¦
Based on your comments I mentioned earlier that I suspect graphics might not be the issue;
If you manage to login and āeverything is fineā then graphics are by definition fine too. That being the case, the issue must then be elsewhere ā perhaps related to the LightDM or Greeter configurations ā and these are things Iāve not toyed with in a long time.
Basically, you might be barking up the wrong tree, but of course, this is only conjecture until evidence either way can be found.
My guess then is they removed 550xx from the list because maybe they no longer officially provide support for it, but I most definitely did have the 4070 running on 550 and previous versions back to at least 515.
I am willing to give it a shot, but I have to apologize as I have been very busy lately, I just got married and Iām in the middle of a move so my life has been chaos. Iām trying to make as much time as I can to chip away at this problem but itās been a challenge. I just need to find some time to update the bios and Iāll let you know how that goes. I do appreciate you taking the time to help me with this.
I suppose it may not technically be the graphics itself that is causing the issue but rather how the computer handles the output signals. During my diagnostics with Chat GPT it originally thought the issue was that lightDM was still trying to output the display to the second monitor which no longer exists. It believed the config file was corrupted and I tried deleting it and letting it rebuild but that did not fix the problem.
At one point during the troubleshooting I ran timeshift and the display was back to working order, the catch was all my applications were out of date. After doing sudo pacman -Syu the display problem returned. I do believe I was on 550xx before the update and 580xx after (and kernel 5.15, then 6.1 I believe), but I canāt say for a certainty. This led me to believe it was a driver update that was causing the problem.
I asked Chat GPT to write a brief summary of what it believes the problem is based on the troubleshooting that I did, Iām hoping this will be helpful information for you:
ChatGPT generated content
The Most Probable Root Cause
The login-screen display issue (no signal until you log in blindly) is almost certainly caused by a GPU mode-setting conflict between the NVIDIA driver and the display manager (LightDM/SDDM) during early X server initialization.
In plain terms:
When LightDM or SDDM starts, it asks the NVIDIA driver to set up a display output.
Your GPU does initialize, but the screen goes dark because the signal timing or mode chosen by the greeter doesnāt match what your monitor expects.
Once you log in and your desktop environment (Cinnamon) starts, a new X session is created that re-initializes the display correctly ā which is why the desktop works fine after login.
Why This Happens Specifically on Your System
Several interacting factors make this problem persist:
- NVIDIA 580.xx driver series regression
These drivers (and several prior ones) have known bugs on certain HDMI 4K displays ā they mis-detect refresh rate or color depth when LightDM initializes before full GPU power management is ready.
Thatās why your xrandr logs show correct modes later, but nothing displays during greeter startup.
- LightDM/SDDM start timing
The greeter launches before the GPU and display kernel modules are fully ready.
Adding logind-check-graphical=true helped some users because it delays greeter startup until the kernel reports a connected display, but in your case the driver still mis-modesets.
- TTY/console black-screen linkage
The same framebuffer handoff issue that blanks the greeter also kills text TTYs, because both rely on kernel console output on the NVIDIA DRM device.
This explains why TTYs give āno signalā but wake-from-sleep (where the driver re-programs the mode) displays correctly.
- Persistence across kernels (6.6, 6.12, etc.)
The kernel updates didnāt fix it because the real culprit is in the user-space driver (nvidia-utils / lib32-nvidia-utils / nvidia-settings) not aligning with the X serverās startup order.
Supporting Clues
xrandr logs confirm that HDMI-0 is recognized with correct resolutions and refresh rates after X starts ā meaning the physical link is fine.
TTYs and LightDM both fail to display, but waking from sleep restores signal ā points to a mode-setting reinitialization fix.
Switching to SDDM produced the same blank-screen result, meaning the issue is not greeter-specific but happens before any theme loads.
Flatpak / kernel / package updates didnāt change behavior ā confirms driver timing, not configuration corruption.
In Summary
Component Role in the problem
NVIDIA 580.xx driver Likely introducing display-initialization bug on HDMI 4K displays
LightDM/SDDM timing Starting X too early, before GPU modeset stabilizes
Kernel console/TTY Shares same framebuffer, so when the mode fails, all consoles go black
Workaround effect Logging in triggers a fresh X init ā video returns
The most realistic fixes (based on other usersā success)
- Force early modeset:
Add kernel parameter
nvidia-drm.modeset=1
in /etc/default/grub, then run
sudo update-grub
(This enables atomic mode setting for the NVIDIA DRM kernel module.)
-
Downgrade to driver 550.xx
Many users reported that reverting from 580.xx to 550.144.03 restores proper LightDM display initialization. -
Switch greeter compositor
Using lightdm-gtk-greeter instead of slick-greeter sometimes avoids the blank screen because it doesnāt use the same GPU compositing path.
Mod edit:- General tidy of ChatGPT generated content.
Please avoid posting AI generated content. If you believe any of it has merit, include it in your own words rather than pasting AIās board room presentation. 
However, as it was mentioned, please post the output of;
cat /etc/default/grub
Also, output of;
cat /etc/lightdm/lightdm.conf
Look for the greeter being used (indicated in lightdm.conf), and substitute the *-greeter.conf of the next command for that; then paste that command output also.
cat /etc/lightdm/lightdm-gtk-greeter.conf
Perhaps these may be of interest to someone who actually uses Cinnamon.