What he means is to use the Render Layers function in the Render Globals. This takes a little planning, as you'll have to render in successive layers, using alpha maps, so that you can bring them into a compositior (e.g. photoshop) for your final image. You kinda have to visual the backgound to foreground structure of your model.

There's no reason why you shouldn't be able to render that scene though. I'm using only 2 gigs of Ram, with a dual-core processor, and I've managed to render all my more complicated scenes so far. You probably just need to adjust your modeling to use less polys in your base mesh.