I'm checking on the state of the art for more-than-two monitors.
I've got some Dell workstations and lots of monitors. They have the NV420 Quadro video card. The output from the card is DP, but I asked for DVI connectors, so they gave me this neat "two headed" dongle, in each end of the dongle there are 2 DVI plugs, so I can connect up 4 monitors altogether.
In MS Windows (what Dell provides, what my students like
), we get more or less "automatic" use of 3 or 4 monitors. We have to fiddle around with which monitor is plugged into which plug, but in the end we have a desktop that spans 3 monitors. The Nvidia Windows config tool does not let us arrange the monitors at all, it is rather stubborn. But we can see and use them.
In Linux, I have more trouble. I have set up twinview with one of the two-head dongle plug thingies, and those 2 monitors work together well. I can also run a separate X server on the other video dongle thingie. I have followed the documentation and many of you seem only able to start a wholly separate X server.
From watching the Windows thing 'just work', I know the hardware is capable of seeing all this monitor space as one desktop/viewport/workspace (choose your jargon, I don't see a difference)
I'd like to fix the Linux side because:
In use, the two separate X server setup is BAD because programs can't work together across all displays. The Gnome desktop/panel does not seem to understand the fact it is interacting with two separate X servers. So sometimes you hit the panel to create a program on the 3rd monitor, the one that is on a monitor and X server all by itself, and that program opens on the other Xserver. I know, it sounds impossible, but it happens.
I have found some posts here from last year about ways to "cheat" to make the 3 monitors work together in 1 X server. The essence seems to be to make one device with Twinview, and then use something else to knit the other monitor onto that collection. But I can't exactly understand how this would work, so I came here to ask "what is the state of the art in 2010"?
Center for Research Methods
University of Kansas