Why can't AMD develop something in-house, and save a few bucks on integration? When the fantastic number NVIDIA is willing to shell out appears, you'll know what I mean. Besides, the laws of physics simulated are well-known, and it's hardly anyone's IP.

I can't help but think that dedicated physics hardware is an expansion card too far. Surely better to use a core or two on your quad core box for something other than checking you mailbox while you play Team Fortress 2. Of course, Havok will now be optimised for Intel, and while making Physx "processor agnostic" might make friends in the developer market it would also help out AMD rather more than nVidia would probably like.

Hrm, any bets how soon nVidia will either fold the functions in their next GPU, or simply kill it off "possible dangerous competitor" (which they happen to have done before .......)

Not that I really care, I'm one of those idiots that thinks games should PLAY right first (no tons of patches before it functions) and THEN focus on making it look "right" ....

On paper they have great cards, unfortunately the reality is different. There's currently no sensible competition to the 8800GT.

Nvidia will fold the Ageia stuff into their Tesla high performance computing products and include some of the functionality in their desktop cards. Given that their roadmap is looking extremely boring (DirectX 10.1 - where art thou?) they need something to spice it up..

Not another one! Open your eyes and do some research... the HD3870 is a much better card at a lower price than the 8800GT, if you only take the time to set it up correctly. If you want your card to run fast and give you fantastic image quality (noticeably better than Nvidia) you should be looking at one of these cards. Granted, the 8800gt has slightly better frame rates in things, but the difference is marginal, and the image quality is much lower. With AMD, at least, they are developing quality cards, not just quick ones.

If you want to read about it, go to the MicroMart forums and search for posts by Slippy, who has done quite a bit of research into this. There are many others elsewhere, but that is the only one I can remember off the top of my head.

Ageia bravely tried to carve a niche but they were pretty quickly gazumped by physics-on-GPU. I suspect NV were more interested in hiring the expertise than acquiring any hard assets. Proprietary physics APIs must surely give way to open standards the way Glide succumbed to OpenGL.