First draw your entire world from the perspective of this camera, into a rendertarget.
Then draw the world again, from the perspective of the main camera, into the main backbuffer, and map that rendertarget texture onto your 3D object that represents the screen.
This obviously gets increasingly slow the more such screens you have, as you can end up rendering the world many times over!
XNA Framework Developer -
blog -
homepage