The best thing about using 3D, is you have so many options.
For instance, you can make a tileset map as a big 3D object, as lots of plains, as a matrix with each tile textured to suit, as a memblock object. A matrix is actually a pretty handy way to work, easy to get started.
If you position the camera above a matrix and rotate it 90,0,0 then it is pointing straight down, so then the 3D X and Z coordinates become the screen X and Y coordinates. You have to decide on a world unit size, like if your tiles are 32x32 each, maybe just using 32x32 for the tile size is the best bet, emulate a 2D coordinate system that way. Matrices are good for tile sets, they use one as a standard, so you just really have to worry about setting the matrix tiles, then position the camera on top, pointing down, and you have a 2.5D tile engine.
Most of my games are fake 2D, I just prefer to work like that. My current project uses plains, big backdrop plains and foreground plains, plains and alpha trans textures for particles. I have a segmented plain mesh as well, that I modify with vertex commands to make an animated water effect, realtime water, even the UV mapping is realtime.
With 3D, it's not a case of ignoring one dimension, you can never do that because there is still Z depth sorting and other issues like that. It works both ways - 3D can introduce problems with zdepth, but it also introduces shaders, bone animated models, different effects for particles etc etc.

Health, Ammo, and bacon and eggs!
