How do you validly compare B:AA performance with PhysX unless they allow AMD to use GPU accelerated PhysX? Or are you saying that even with PhysX and all other settings the same, it will be 20% faster still? Therefore given PhysX load, it will actually be considerably faster without it, in which case wouldn't they use the faster value as a comparison? Unless somehow they now get PhysX processing for free?
AMD doesn't have PhysX and PhysX will be part of GF100 so (for nvidia) it's a great comparison. As for considerably faster without it, that's what I'm guessing against. PhysX hits NV cards with about a 30% hit and AMD cards (i.e. CPU-based PhysX) with about 80%.