Xinerama of course still has its use, that’s why I asked “what’s your use-case”.
In most cases, users throw some gpus in their system and expect to them just adding up. Nope, doesn’t work. You can’t throw 10 gpus in your system, connect 20 monitors to them and get the 10 fold performance out of it.
Then there’s the problem with Xinerama on nvidia proprietary driver and Xinerama itself. Since it’s not widely used anymore, it’s in a state of decay, many things/setups broke starting with xorg 1.16/1.17.
Doing Geforce nvidia+nvidia isn’t really the way to go, if you’re into multi-monitor, just go for Quadros and use Mosaic. Otherwise, use Geforce+intel or Geforce+amd and use PRIME. Once nvidia implements the output sink feature in their drivers geforce+geforce will also be usable but don’t hold your breath waiting for it.
Quadros are priced for insanity so that will never be something I feel OK with. Frankly I think the GPU market in general is brutally over priced in general and the “appeal” of Quadros are things I don’t really care about when the generic GPU’s support 4+ outs and should work. I think the better way to look at it is why the hell isn’t Mosaic supported on all GPU’s. I know their rational for this but I feel it just a cash grab crossed with justifying product segmentation.
I actually tried AMD+nVidia before, no go. Tons of other issues where as 2 nVidia with the proprietary GPU’s actually worked out of the box…more or less. One of the other “PITA Gotchas” however was the fact I had 6 monitors all the same make/model. This caused some unintended fun when they wouldn’t be told apart for some configs, but I got it working eventually.
Another unrelated fun problem was my original 6 screen set up has Xinerama enabled for one GPU, not the other (separately) but then I had xinerama enabled for both. So one set of screens was seen as a single, while the other GPU saw 3 separate but allowed me to use all the pixel space. It wasn’t until I dug through my xorg.conf that I realized the nVidia drivers were not reporting a duplicate xinerama line that grouped screens on one GPU, then all the screens across both GPU’s.
Anyhow I’ve yet to readd the second GPU and test with POP_OS…been a nuts day.
Quadros are PAL - pay the additional luxury.
But you reminded me of a common Xinerama setup to work around problems, using a ZaphodHead setup: every head has its own screen https://download.nvidia.com/XFree86/Linux-x86_64/440.64/README/configmultxscreens.html
which are then tucked together using Xinerama and a fixed layout. Did you try that?
Both GPU’s are pretty much a screen per head, that’s the point, I’m not running 6 GPU’s, just 2, one can do 4 screens, the other 5…so long as I don’t hit the pixel limit.
Yes, the horizontal layout requires I configure 3 XScreens on 2 GPU’s. Long story short I no longer have my 6 screen set up. Several were defective so I sent them back. So the idea (for now - until I can find suitable replacements) is/was to readd the second GPU, run my center screen off the 1660 so any game I play has the better GPU while the flanking screens run off the 1060. Initially this caused a strange problem because while I can set this up with 2 XScreens how the layout sees it is the center screen is “covered” by the generic area (despite being empty) from the 1060’s XScreen. So I have to make a separate XScreen for each screen on the 1060.
Well tested with POP_OS nVidia but gdm doesn’t understand what to do once I enable xinerama. I actually had that issue before too but I can’t remember what I did to fix it since it was months ago. Feeling rather burnt from this stupidity. I mean it’s just a handful of applications that have this idiotic offset issue but why? Thinking it would be better to toss the damned GPU in another machine and just make “fake” screens on my workstation then have the other machine remotely display the fake screens from the work station on the physical screens connected to what should be the “second” GPU.
Just tested configing my center screen to be the left most, XY issue gone, unusable but clearly something in the way either the drivers or how X sees things is giving the wrong XY because of the “gap” between 2 screens that it thinks are out of order…