@ efd753: I see what you're saying, why can't the dual-core CPU architecture handle these kind of physics calculations instead?. Well, the answer lies in your own question.
The idea of us having a dedicated GPU is just that: a dedicated processor that can handle all your eye-candy calculations that a) the CPU isn't (or no longer is) designed to do, and b) even if it technically can, the GPU still offloads the tasks from the CPU to improve overall system performance.
If you watch the tech demos, you'll get the drift of what a PPU is capable of in games/real-time. There's no way in hell that current CPU's could produce this level of realism, parallel architecture or not.
AMD X2 4400+ Toledo 1MB/Asus A8N-SLI Deluxe/2GB Kingston HyperX/SLI Gigabyte 7800 GTX 256MB/Creative SB Audigy 2 ZS Plat w/ Creative Inspire 6.1 speakers /180GB Hitachi PATA/19\" AG Neovo F-419 LCD/OCZ PowerStream 600W PSU/WinXP Home SP2