Display walls (compound flat monitors with a very large number of pixels) is a very useful tool for geovizualization because of their unique ability to display multiple geographic scales at once. My paper on the topic:

Sorokine, Alexandre. “Implementation of a Parallel High-Performance Visualization Technique in GRASS GIS.” Computers & Geosciences 33, no. 5 (2007): 685–95. doi:http://dx.doi.org/10.1016/j.cageo.2006.09.008.

Also display walls have a very strong Wow! factor if you are interested in that.

Displaying PowerPoint slides is possible but not the goal, good projector will be much-much cheaper, better, and easier. Even having such capabilities may be a disadvantage because people will use the system (and waste your time) for the purposes it was not built for.

The main part of a display wall are monitors. You can build a display wall with any kind of monitors but you'll get better result with monitors with narrower bezel, larger display area, and higher resolution. The main limitation on the number of monitors is how much physical space you have, for really lots of monitors electric power may be the limit. You should keep one additional monitor for the head node. If you have a choice choose the monitors without glossy screens or other glossy surfaces.

Using flat-screen TVs instead of computers monitors is also possible but TV tuners some time cause interferences when TVs are are put next each other.

Monitor connections should fit graphics cards connection. There are several types of connectors in use:

• VGA -- Good old connector but supports limited resolutions (no 4K) and limited cable length. Not used in newer monitors.
• DVI -- Also an old standard but I won't call it good, more expensive cable than DVI, limited length, have several subtypes that are not interchangeable.
• HDMI -- commonly used connector in the TVs, has built-in DRM that may cause stupid problems with some graphics cards, do not use the cheapest cables.
• Display Port -- most up-to-date connector, may not be available in older monitors and graphics cards.

Monitors have to be arranged in a regular grid that is mostly flat or slightly caved with minimal space between monitor edges. The simplest implementation is to put lower row of monitors on the floor and upper row of monitors on the chairs. However, much better solution is to build a DIY monitor mounting grid out of aluminum railing (better) or any other material that is strong enough to hold monitors. All modern monitors have VESA mounting holes on the back and they can be used to attach a monitor to VESA mount brackets (~$10 each + $5 screws). You can buy commercial mounting harware for the walls but its pricing is comparable to monitors themselves.

If the wall is big enough it can be made slightly curve to improve readability by the peripheral vision and reduce glare. Strongly curved walls are of lesser use for GeoViz.

Monitors, especially old models, may take lots of power. You will need good quality power strips. Also it is a good idea to have some kind of equipment to power monitors on/off all at once.

Three setup options:

1. Fat node: one powerful PC runs the whole wall. Pros: easy to setup, [most] applications will run without modification on the wall, no need for the head node, you hae to maintain a single PC. Cons: pricey ($15-20K for the computer and graphics cards), number of pixels is limited to around 30K, some applications like Google Earth won't run at full resolution because they do not think that such huge monitors may exist.
2. Cluster: several PCs each running one or few monitors plus one head node and (optionally) file server, software takes care of distributed rendering. Pros: can use any trashy PC, practically unlimited number of pixel, easy to expand when resources become available. Cons: needs lots of power, takes time to setup and maintain software, most of the software will not be able to use full resolution without modification.
3. Video processor: the dumbest, most useless, and most expensive option, does not make any sense for GeoViz: monitors are connected together via a box that will scale video signal (from a PC or other source) to multiple monitors. Pros: there is no difference with using regular monitor for any laptop or PC or any application. Cons: resolution is limited by the resolution of the PC graphics card, very expensive (~$80K).

For the cluster you can use any number of any computers that can run Linux (i.e., almost any computer). It's good to have identical systems but it's not a requirement. The number of PCs you need depends upon the number of monitors you have, i.e., total number of video outputs should be no less than the number of monitors plus you need one box as headnode with its own monitor that is not a part of the wall. If you have more PCs than you need for the monitors you can also setup one node as a gateway to external network (has to be setup for NAT) and one or more file servers.

Reasonably fast network switch or hub, can use any standard hub but faster the better. If you PCs have more than one network interface you can create two networks seprating video traffic from other traffic. If your sysadmin is experienced enough, you can setup network boot for the grphics nodes that will save lots of maintenance time.

Each node must have a standard Linux install (Ubuntu or Fedora or whatever distribution you like) with XWindows and ssh. It should be possible to login from the head node to graphics nodes without a password.

The old way is to run Xdmx on the head node. This way you will be able to run [almost] any standard XWindows app on the head node and its content will appear on the wall in the full resolution. If you succeed to hook Xdmx up with Chrome OpenGL renderer then you will even be able to run OpenGL apps. However this method is slow and outdated.

To play movies or display very large raster images you can use Blockbuster (free application) or commercial application enVideo.

The right solution for GeoViz (for 2016) would be to customize some WebGL renderer specifically for GeoViz needs. It will require some programming but the result will be much better than with the standard tools. One idea of such application is to use OpenLayers or Leaflet with Electron or NW.js and pull all the necessary geodata from web services.

There are several applications outside of GeoViz domain that support display walls for general scientific visualization, for example, Visit from LLNL and some other VTK-based systems.