I'd have thought it'd be cheaper to put a second camera on the other side of the mirror, in the next room if you like, and make the mirror semi-transparent from the other side. Saves you duplicating all the vertices (especially all those in complex player characters), and saves from doing twice the vertex shading I think. You still have to two twice the number of final perspective transforms, but I'd hope those are pretty well optimised in modern GPUs.
Although more often it's just a duplicate of the room with your character model running around, mirroring your input.
More impressive is some engine's ability to display another area of the map in realtime on a monitor.
A monitor in another room could just use the same trick, but with the second camera in a different entire model. You don't get any of the savings I've mentioned above then, but then shading requirements for an image displayed on a monitor in game are probably quite different from the normal ones anyway, and it's likely you can get away with a reduced framerate for that display anyway.
Edit: Not that I know how modern GPUs affect 3d rendering. Last time I looked into this was a good while ago.
Edited by mal at 18:06:11 07-01-2013
Cubby didn't know how to turn off sigs!