If you're going to do live input processing, you're always going to have that little delay that messes you up, no matter which OS you're on. In fact, the only good way to do this is to do a scratch mixdown of all your tracks and start a new project just for vocals. That way you can record your vocal track with only one other track playing, and probably be good at < 5ms latency, which you might not really hear (especially when you're talking about reverb...just add the 5ms to your predelay)
Once your track count starts climbing, you're going to have to crank them buffers up, and live input processing becomes much more annoying.
If the alternative solution to this problem is a lame debate over operating systems, then I'd suggest getting an outboard reverb unit.
Slackmaster 2000