lets clear this up. The ATI system is completely different in architecture and implementation than the Nvidia system. This is comparing apples and oranges again. ATI has always had an unusually large number of processor cores compared to Nvidia. In this case more does not equal more power. There is a lot of other stuff going on the chip level than processor core count.
Good example is a ATI GPU core does physics better than the Nvidia core of the same class; BUT the Nvidia card will have an Aegis physics chip that it offloads the physics to which gives it a huge boost in physics. It definitely has a huge number of PPU's that don't get counted when shader specs are listed.
People who correct me often amuse me.
The reason that 240 beat 800 is quite simple. We're talking about a Superscalar architecture. Every 5 ALU's only have a single functioning 'brain'. Each of the 240 are SP's and have their own 'brain', as the architecture is Scalar, which is the next level upwards, just as Superscalar is to Vec4 (Vector - 4 ALU/'brain').
ATI doesn't do physics at all, actually. But, if they did, the performance would only be relative in the exact same way as real-world performance. Or, if nothing else, nVidia would have the upper hand. As I've said from the very beginning of G80 - Scalar is better and Superscalar is out-dated. It's not that ATI has 320 and nVidia has 128, or 800 and 240 and whatever comes next. It's that 1600 fail to beat 480.