by horsawlarway on 6/26/2025, 12:00:39 PM
by encom on 6/26/2025, 11:48:35 AM
I've been running a mixed DPI setup for a long time now, with a 4K display center, and a vertical 1080p display on either side (so 3 displays total).
My conclusion from running this and similar mixed DPI setups over many years is that mixed DPI is extremely poorly tested, if at all, by all vendors. KDE Plasma on Wayland finally pretty much gets it right 98% of the time. X11 wasn't a great experience, regardless of what's technically possible.
Windows has so many annoying issues. Apps are often blurry on the low DPI displays. The mouse cursor has no concept of screen DPI, and treats the entire working area as having the same DPI, so you have to hit the "exits" of the high DPI display just right to land on the low DPI ones. Positioning a window across different displays only scales it correctly on one. There's probably more, but I've been 100% Windows free for a couple of years now.
I wonder how OS X handles it. I don't like that OS, but it sounds like the kind of thing Apple would care about getting right.
by baobun on 6/26/2025, 12:13:47 PM
TIL the major reason mixed DPI is subpar on X11 is because Gnome is blocking support on Gtk.
by the_mitsuhiko on 6/26/2025, 12:04:24 PM
I'm not sure where all these "X11 is actually great" posts recently are common from but X11 does not have DPI solved. I encourage you to read this comment [1] to better understand what the actual situation is like.
[1]: https://lobste.rs/s/ceylzx/forbidden_secrets_ancient_x11_sca...
I think this post is one of those "technically correct" but functionally wrong rants.
My experience with multiple monitors at different dpis hits nearly every case of failure he points out.
It's a lot more work to configure.
Apps fail to account for it.
Spanning two monitors results in terrible scaling problems.
Apps that do account for it at start up won't account for it during reposition, so they look fine if they open on the right monitor and terrible if moved to the other.
Getting solid workflows for flexible positioning requires hacks like mentioned at the end for xrandr.
Etc...
So sure, you can do it and it sucks.
That's not really the win I think the author seems to think it is.