Considering the price ranges they were designed for consoles had amazingly good specs for the time, but I digress.
I've never disliked PhysX - I've wanted a lot more games to implement it. Mirror's Edge was even more awesome with PhysX.
I'm hoping that CUDA and ATI's equivalent (forget what it's called) help with this more than they have so far. It's not so much the calculations themselves that chew through processing time but the fact that they need to be processed serially, which considering the number of objects needed for any realistic physics simulation will bog down the average CPU as every object needs to "wait it's turn".
Hell, I actually think ATI would perform better on the physics simulation side than nVidia (and I'm an nVidiot
) simply due to the fact that their cards can handle a greater number of parallel calculations, and for physics the processing power isn't as important as simultaneously calculated objects. There would still be overhead since things would still need to go through the CPU, but far less.
I'm betting you ATI's gonna grab up Lagoa, once both major companies have their own physics software I think (PC) game companies might actually start taking greater notice of physics in games (aside from ye olde Havok ragdoll).
And for anyone who wants to dispute the advantages of parallel processing of many relatively similar calculations... try the CUDA version of F@H or hell, grab Apophysis and Flam4Cuda and compare rendering times on CPU and on GPU. You lose out a bit in precision but that's being worked on slowly but surely.
And then I could just be talking out of my gat but I needed to get all that out