Just a quick noobie question: Ive heard alot about physics in games lately, ive heard they have created a physics chip and most game have physics engines. Could somebody please give me a few hints at what you would include in a physics engine? Cheers, Nick

I think it's safe to say, we consumers won't be seeing physics cards for quite some time; it is a new technology, after all.

As per your 2nd question, collision detection is a must.

http://www.extremetech.com/article2/0,1558,1777187,00.asp

This page talks about the physics card in depth.

The only problem with the new physics chips are going to be the fact there is no standard API for them yet. I am sure that the industry will standardise at some stage but until then it will be a bit like the old DirectX and Glide API battle. And this time i'm sure that M$ or similar will likely win again.

Like the GPU, the new PPU definitely will free up the processor, but as said before, will require an entirely new API.

Quote:Original post by visage
Like the GPU, the new PPU definitely will free up the processor, but as said before, will require an entirely new API.

Novodex is not a new API.

Ageia is making the chips. Ageia owns Novodex. Novodex is an established physics library. Novodex will use the PPU if it's there, or not if it isn't. I don't think there will be a way to access the PPU apart from through the Novodex API (but I may be wrong).

http://www.ageia.com/

Some people are of the opinion that it will only be useful for "eye-candy" physics - it will still need to be possible to run all the physics essential for gameplay in the absense of any PPU. For this reason some people think it isn't going to be half as interesting as its proponents claim because it won't actually open up gameplay physics. It's hard to imagine a game getting released that _requires_ a PPU.

As for what goes into a physics engine, well have a look at what Havok, Novodex, Meqon, Trie Axis, ODE, Tokamak, Newton etc provide. Incidently, collision detection doesn't have to be part of a physics engine - it's nothing to do with physics (though certain implementation details will be tailored to the physics implementation). A friendly physics engine will provide a default collision detection package, but a really nice one would let you extend it or even swap it out for your own package. Some physics simulations might not even need any collision detection at all - e.g. a boat floating in the middle of an ocean doesn't "collide" with anything, but there's still a whole lot of physics in the interaction between the boat and the waves.

Quote:Original post by MrRowl
Some people are of the opinion that it will only be useful for "eye-candy" physics - it will still need to be possible to run all the physics essential for gameplay in the absense of any PPU. For this reason some people think it isn't going to be half as interesting as its proponents claim because it won't actually open up gameplay physics. It's hard to imagine a game getting released that _requires_ a PPU.

Same problem as in the early days of 3D then. :)
Back then, they also had to support software renderers, so they couldn't actually do much new, only make the usual stuff look better.
And the PPU will have to start out the same way. Of course, if they can make everything look *lots* better, everyone will buy those cards, and then they can start relying on people to have them, and start doing really fancy things.

Does this mean that from the point the PPU is a common thing in
every pc has and the API for this chip is stable and running that
programming physics in games gets a lot easier ?
I mean do a physics programmer still has to know math to do
physics in games or is the API going to take care of it?
If this is the case there are some boring times ahead for physics programmers.

That's like saying there are boring times now for graphics programmers... Im not saying it's not true, just comparing :)

I suspect that the chip will find itself into either a play station or an X box. At this point, the other side will be forced to stick one in theirs for marketing reasons. Console games will then be able to rely on it, and the desktop computer will follow suite, possibly with a software emulator for it along the same lines as mesa for OpenGL.

Seems a shame to specialise the thing for physics though - it looks like yet another parallel processing unit which we can't quite use for anything we like - I'm hoping that one day we will just be presented with 10,000 processors all doing SIMD and the world will open up before us.