Nvidia following with Physics Simulation on GPUs

Aegia will have an advantage in that nearly all ps3 ports will need the ppu to play.

 

I don't expect Ageia to last either, but for other reasons.

As for advantages/disadvantages of different solutions, let's just wait and see, shall we? This is all a bit of smoke and mirrors still, all I've seen so far from both nV and ATI is PR bull and a couple of worthless flashy demos.

 

I am not sure if Ageia will last in its present form. In fact I think it will attempt to get success and an installed base, split and go public then sell to the highest bidder.

Doesn't mean that hardware accelerated physics isn't here to stay.

 

Well I see physics on GPUs as a wonderful way for the graphics vendors to obsolete us. Now not only will graphics overwhelm the card, but also physics calcs. And don't doubt for a sec that this physics stuff basically maxes out the video card. Hell right now the fastest cards can't handle the shading workload. Do I want future cards to have less graphics performance but do physics instead? Not really. Separate card is nicer in many ways IMO.

 

Yes, especially because their PPU is probably pretty much like half a Cell. The PS3 devs will do the hard work and define such a new Physics/Scene management API, if anyone will.

 

Whether the PPU is anything like Cell is irrelevant when you aren't allowed to program it.

 

What do you mean with "not allowed to program it"? There surely will be API/drivers out there

 

Right now, the only way to program for a PPU is by using a very high-level physics API, which would be akin to programming for old fixed-function OpenGL. There are no "shaders" available for a PPU.

 

Yes, but they need to think of a way to do (and communicate) meshes and scene management. The PPU API's are rather "dumb" at the moment.

 

found ATI's responce the most interesting. I think the "Direct Physics" API will problably make or (most likley break) ageia. 3 Different and exclusive API's for hardware + X software ones wont fly for dev implimentation in anything but whoever can shove the most money at what ever game title. I'm going to expect Ageia to make as much money as they can off their solution, and then once Direct Physics gets implimented and noticed, Ageia may try to (for lack of a better word) whore themselves out to Nvidia/ATI hoping to get picked up by one. When it comes to competition and talent, Ageia has no chance against either company.

 

Microsoft APIs don't always win, especially very high level APIs. Remember D3D Retained Mode? Remember DirectEngine? Why does Miles Sound System still exist? What if DirectPhysics is missing a particular simulation effect, or Microsoft going to release upgrades as fast as Havok and Ageia? Is MS going to give out the source code to DirectPhysics so game companies can tweak them like Valve did with Havok?

 

Now this is nice, first look into the architecture of the PPU I've seen:

http://www.blachford.info/computer/articles/PhysX2.html

 

I suppose when they said their architecture was very similar to Cell, ...

/quick skim

 

Wow, yeah, that's pretty much exactly like Cell. Since I fully expect CPU manufacturers to go the way of Cell within a few years (3-4), I'm willing to bet that this PPU will be utterly redundant.

 

I am wondering whether it might be an option to put Cell on GPU cards to function as a PPU and maybe also assist the GPU with some things when not used as a PPU (eg. vertex shading) and do mpeg encoding/decoding, sound and other things like DRM decryption for protected movies as well. The Cell chip is going to be manufactured in huge quantities and so it will be relatively cheap compared to the custom Ageis chip.