Since the density gradient method is intended to let the macromolecule reach its equilibrium buoyant density, the spins should be long enough to achieve this. I have always done overnight spins, 18-24 hours.
Dear Luke, Adam is absolutely right for the time scale of the experiment. But the most important point is also the centrifugation speed to apply, so that you effectively reach an equilibrium without pelleting all of the protein (as in sedimentation velocity experiments). The right speed to apply is dependant on your protein complex mass, and usually for good results it is important to make different runs at different speeds.