OK, I've run a similar test using, of course, a different model (I wasn't about to try and copy AgentSmith's gorgeous model) ;). Both samples were rendered with fast preview mode on, quality set to default (no AA) so as to get the real jaggies to come out. Sample A was a straight render at 404x404 at 72dpi. Sample B was rendered to disk at 808x808 at 96dpi. Then it was brought into Photoshop and resampled to 404x404 at 72dpi. The differences are blatant and quite obvious to the naked eye. If anyone would like to conduct the test for themselves and would like a copy of the model (3D Studio Max 3ds format) used in this test, just drop me an email.