Anyone using three 4K monitors or three 1440p monitors?

I run Qubes with three 1080p monitors without any issues on an AMD GPU. I’m looking to upgrade to three 4K or 1440p monitors.

There is a Github issue for improving support for high resolution displays but I can’t tell how severely the remaining issues affect daily usage.

If you use three 4K or three 1440p monitors with Qubes, can you share your experience? Do you experience lag, tearing, choppiness, etc? Are you able to watch videos in a qube? Can you full screen applications without issue?

1 Like

I’m using 3 x QHD (2560x1440) displays, it gives a 7680x1440 desktop, it works without any issues using AMD internal graphics.

I don’t know if 3 x 4K works, but the resolution is so high I don’t know if it works with internal graphics, they normally max out at 8K@60Hz

1 Like

I have two 4k monitors but have my display resolution scaled down to 1440. If I did it again, I’d just buy one ultrawide monitor to save space on my desk and skip 4k.

Regarding 4k vs 1440, nothing scales well or consistently to be readable at 4k. You can adjust some things but others are either impossible or very difficult. Trying to adjust DPI across templates and every single AppVM sucks. This may be acceptable for a desktop that never moves but is painful for a laptop, as you have to repeat the frustrating adjustments every time you switch from docked to traveling. The adjustments between 1080 and 1440 are minor for me.

Qubes OS also has a weird behavior where if you start VMs then expand the display resolution, those VMs don’t know about the new resolution. So, if you move an AppVM window to the new area the mouse doesn’t work. You have to restart the AppVM(s) for them to be useable with a mouse; they will display fine and be useable by keyboard but not by mouse until restarted. Again, not a big deal for a desktop. But frustrating for a laptop that boots at 1080p then transitions to the dual monitors.

2 Likes

What is your hardware?
You may be going over your GPUs maximum it can handle.

I run a 1060 and the maximum for that GPU is 7680x4320@60Hz. So it can handle ~33 million pixels across my displays before it can’t do more and has issues.

So I recommend that you look into the maximum for your GPU, not per display, but overall.

Are you running HDMI or DP? That will also have an impact on what it will do as well.

Please provide as many details as you can.

Also, have you turned off composer and turned it back on?
I find that that removes many issues after install in relation to graphics.
Composer is just bad to have anyway.

Window Manager Tweaks → Composer → Enable Display Composing

uncheck and recheck it.
Then turn off the shadows and whatever else you don’t want.

just unckecking it to disable, and checking it to re-enable fixes a lot of display issues.

1 Like

Yes, I have a three 4k-monitor setup. It was the mother of all headaches to get everything working acceptably well. However, what I have now resembles the stock gnome installation with 2x scaling I used to run previously with Arch. This kind of setup produces a few issues:

  1. Out of the box, nothing is properly scaled, and the mouse speed / acceleration are off. This is surmountable, but a LOT of dom0 and appvm config files need to be tweaked.
  2. Graphics performance is barely acceptable but adequate for a productivity-oriented machine. Scrolling through text is not “buttery smooth” - think Windows 95. To make web content work at all, I need to pass through the onboard dGPU to the internet qube for use with virtgl. Which leads me to the next issue:
  3. The plumbing on my laptop routes all external graphics output through the dGPU (even when piped through the usb-c/thunderbold ports). When the dGPU is used in headless mode for virtgl, this prevents external monitors from being driven directly. To solve this issue, I use a Razer Core X eGPU, which is connected over thunderbolt. This is inelegant and inefficient, to say the least, but it gives me 3x 4k tear-free graphics. If you go down this road, I suggest an Nvidia card for the eGPU. On my machine, the AMD driver assesses the cpu/thunderbolt/pcie link to operate at pcie version 1 speed (maybe it doesn’t interpret the speeds reported by the TB chips correctly? Or perhaps it misinterprets the idle speed as the maximum speed?). The resulting performance is inadequate and afflicted by bad tearing. This is a known issue, and there’s a simple if statement that could be patched in the driver, but rather than put up with the hassle, I simply bought an Nvidia card.
  4. Having three monitors increases the the number of security domains you can work within simultaneously. This proved to be more of a cognitive burden than I expected.

DM me if you’d like the script I wrote that automatically configures a new qubes install in the manner described above.

Edit after reading the previous replies: I also wrote a systemd service that looks for the presence of the eGPU card at boot and selects the appropriate xorg configuration file (3 external displays or 1 internal display). This eliminates any hassle when traveling.

1 Like

I’m not trying to disagree or argue. I just think, this approach is worth taking in consideration. I feel like people might be kind of stuck with the idea that two or more monitors is always the best solution, and maybe it still is sometimes but maybe it’s also worth challenging that idea a little bit.

Why would you want borders and gaps between monitors when you could have same or larger screen estate without and there are plenty of ways to divide the screen programmatically?

TLDR: If your job is to write code all day or stare at Excel spreadsheets, buy an 8K TV instead of a multi-monitor setup. You can even use the same TV for 4K 120 Hz gaming or watching movies as a bonus!

1 Like

One reason would be GPU pass-through, where you need to dedicate a screen to the GPU. I want to be able to use one screen for the GPU, while still being able to watch what is happening in my running applications.

1 Like

You are using Gnome? You should be using Qubes?
With 2x scaling you are running 8K on each monitor? Can your GPUs handle that?

Out of the box nothing would be scaled correctly on a 4K monitor because of your settings.
Just set up the displays corretly and it works. I have to do this with Linux often enough. Takes a few moments.

You would need to assign more VRAM to your guests then. Depending on the operating system you are having issues with. I have scrilling buttery smooth on my system, even when running 4K.

You can use just the integrated GPU for all displays, even the external ones.
Not sure on the method that would be used to set that up on your system, but I have done it before, over a year ago, so I’d have to look up what I did again, I’d have it somewhere around here on one of my backup drives.

What is the eGPU?
My AMD card didn’t have that issue, everything still registered as PCI-E v.3 .
So I didn’t have an issue with that. Might be something to do with your machine.

I have 3 monitors and work with 15+ in general at the same time. But I have maybe 30+ running at any one time.

Would be an interesting thing to have for laptops of that sort. An addition to put into Qubes for those that do require it.

I jsut use the one configuration file myself thoguh, when changing monitors the system just generally alters my displays accordingly, it just turns them off and on. But remembers their positions and all.
Might just be the fact that I’m using default installation of Qubes though.

I wouldn’t suggest 4K as dpi scaling is annoying on X11, especially with Qubes but I use 4 large 1440p monitors and it works fine with no scaling. (edit: I run the monitors on integrated AMD graphics so I can pci passthrough the dGPU)

Currently I’m running Qubes with Xfce. Previously I ran Arch with Gnome at 2x scaling. Three 4k monitors, both before and now.

Much of the tweaking was necessitated by the use of Xfce, Thunar, etc., and involved not just scaling, but configuration of the greeter, lock screen, mouse, window manager buttons (rescaling by hand), qubes’s videoram, refresh rate, keyboard shortcuts, etc. in order to make the DE resemble default Gnome. By comparison, modern Gnome with Nautilus demand only a few seconds to adjust the scaling to 4k.

As for VRAM, obviously, I’ve given my qubes plenty. But I also run ECC memory, which reduces the clock speed, so that may influence things as well. In any event, if you study the architecture of qubes, you shouldn’t expect performant graphics, especially at 4k.

Whether or not you can use your iGPU in the manner you describe depends on the internal wiring of you laptop and whether (if present) the dGPU is passed through to an appvm. It’s a bit more nuanced than you suggest.

As far the AMD issue concerned, as I mentioned, there’s an issue with how the kernel module interacts with the TB link, and it’s not worth the time to try and fix it.

Okay Qubes with xFCE, good.

There isn’t much required for that when using xFCE. Just set up the displays with the display manager and you are done.

How much VRAM have you given each Qube?
How much VRAM does the machine have?
Does the assigned amount go above the amount assigned to the GPU(s)?

I would recommend you use the discreet only, no integrated.
Will resolve 99% of the issues that is explained.

TB is just USB.

So just use the discreet GPU, and leave it at that. Will resolve the issues.