I'm not so sure about GPU physics on DX11. GPU compute is going to be there AFAIK, but from there to a good physics API there's a long way. DX11 will launch a lot later and taking into account the adoption rate of DX's, we wouldn't have hardware accelerated physics until 2011. No thanks, give the physics now, and I myself will decide if I want it or not.
EDIT: after some searching I have found that Ati said that before Ageia's adquisition. And TBH there is a difference. Ageia had a hardware base of not more than 100.000 PPUs, Nvidia has over 50 million capable cards and 55-60% market share. You can't compare, it does pay off to implement good physics in your game if you know it could mean an advantage over other games when you have so much potential buyers.
They actually said it early august this year:
"in his speech GPG CTO Technology Day held in Iceland’s capital, Raja Koduri, CTO of AMD GPG (ex-ATI), announced that AMD believes that the time for proprietary software solutions such as AMD's own Close-to-Metal and Nvidia's CUDA has passed.
As a result, AMD will throw its efforts behind DirectX 11 Computational Shaders and the OpenCL GPGPU language and will focus on standardized solutions only."
Of course it's true that that is not mentioning physics, and also it's not quite clear if MS will do anything in that direction or offer encouragement or support for physics on the GPU via DX11, on the other hand ATI could use the GPU-computational part of DX11 and write a physics plugin for it, theoretically, as could nvidia, or nvidia could extend CUDA to work on top of DX11 so they can smoothly move their PhysX api to DX11 and have no need to expose their users to big changes in that area.
And then there's the DX11 platform.. it'll be vista/windows7 and not XP, which most people still prefer and have.
So I guess you are right, PhysX seems to be the winner and we'll have to get nvidia cards as assist or someone should make a PhysX driver for ATI.
Personally I pull more towards getting a nvidia card as secondary right now, as soon as I know if that's practical, because some dubious pictures on hardspell isn't quite the same as knowing it works, and works smoothly, to mix cards that is, for all I know it's fraught with all kinds of issues and/or requires hacks and instability I'm not willing to accept.