Another interesting slide we've seen consists of various scenarios for 3-way SLI with a side of PhysX power in a seven-slot motherboard; and according to it, Nvidia is hoping users will go for three dual-slot graphics card and a fourth single-slot card for physics acceleration.

It seems like Nvidia has managed to incorporate PhysX to a fairly decent level into its products by now, although we believe the performance can still be vastly improved with some more time to tweak the CUDA integration.

We're still waiting for the killer application that will make people go "oh, ah," etc. but it seems like we'll have to wait a wee bit longer until this happens. It's also likely that Nvidia will have to wait for people to spend more money on extra cards until this happens. It's the whole chicken/egg effect and no one is really going to care about hardware accelerated PhysX performance until there's a game out there that really takes advantage of it.

However, we're not sure about this whole three graphics cards plus PhysX thing. Wasn't PhysX supposed to lower the processing demands? In saying that, it'll do nothing to help crunch polygons, and maybe that's where Nvidia needs the extra cards so that they can keep on top of AMD.