pouët.net

http://www.pcper.com/article.php?aid=140

The concept is quite interesting although it's mainly aimt at gamers at the moment. I wonder if this could really become some kind of standard like our current GPUs.

Who knows, demos could use this kind of hardware acceleration through a physics API in a few years :)

Interesting. I'd love to take a look at the API though..

I'm sure it will be standard in a few years, quite a few developers are supporting it, and i think it will be very useful when porting between PC and ps3 etc. There was already a physics based demo using a pd physics api (can't remember its name tho :) so this is surely just the next bit of evolution. Is the api they're providing with that card free to use tho?

$200-300 is too much as for now, i think.
well, this is pretty useful for fps and racing games only, right ?
as for demos, i bet even if it will be used, then for everything but original purposes. i mean someone will try to generate textures with it, others will trace rays or even render sound :DD

on the second thought, because of ability of software calculation ( as stated in the article ), i bet those libs are going to be fucking expensive and the whole lib -> hw driver transition stuff will probably be patented, so there wont be a chance of using it for free, at least legally :/

if not then be prepared to see shitload of cubes-and-balls-falling-into-water-or-madel-in-a-waterfall demos :D
along with showed demos ripoffs ;)

Playstation3 have it included

Some engines like that are provided free for "non-commercial use", i.e. you can do game mods etc. without licensing, but if you want to do a commercial game, pay up. Hopefully that's the way they'll go.

I think these chips will have an easier time of being adopted than ray-trace processors.

Of course, a graphics card with integrated traditional GPU, raytrace GPU and physics processor would be a good step to getting both adopted.

An engine with super-realistic physics, and very fast real time raytraced graphics would really rock.

the whole lib -> hw driver transition stuff will probably be patented, so there wont be a chance of using it for free, at least legally :/

I'm pretty sure it will be in DirectX or something eventually. (OpenPL?)

gargaj :
that's simple : if it's free then it could eventually be included into directx ( i really doubt it will tho, but lets see what it comes like in a few years ) or wherever, it doesnt matter where then; if it's not then it obviously will never be... just like that

Very true. While I'm skeptical about it's widespread acceptance within a couple of years, I think somebody will definitely do some *crazy* shit with these.

Gargaj: it's probably the only way to have it accepted by customers and by the whole game industry. I trust those people at AGEIA are well aware of that (hopefully) ;)

Evil: it's NOT part of the PS3 design.

kindof depends how it's used. gpus make a lot of sense and arent at all hard to use, you kick off your graphics in a sequential order and it gets rendered "sometime in future". also, gpus are doing small intensive operations on individual pieces of data and are absuing memory a lot - so again, it makes sense to have that on a dedicated piece of hardware.

for physics stuff tho, it's typically more like you want to evaluate some collision or physics and know it immediately for some ai or controller logic. to reorganise that into some totally parallel processes takes a lot of effort (and gamecoders are often lazy) - one reason that vu0 on ps2 was so underused apart from helping gfx stuff. and one reason i suspect the ps3 cells will be mainly used for gfx, not for as much as they intend. also, a ppu smells like it would be a lot less specialist hardware than a gpu - it would be more like just another cpu and some ram on a card.
take away the gfx onto the gpu and the physics onto the ppu, and you end up with a 3ghz+ cpu doing very little in a modern game apart from a bit of logic and scheduling.

in other words, i'll wait and see how useful it really turns out to be. :)

Actually if it works the way I think it works, it can be really helpful for mesh generation.

FUCK IT

another 300$ shit piece of hardware for what ?
we already have unescesarry complex cpu and gpu that consump too much power

hardware metaballs! \o/

:(

oh wait. that already exits, aight. well whatever, it sucks anyway.

note: the novodex sdk is free for non professional use and can be downloaded here: http://www.ageia.com/novodex_downloads.html

http://www.ageia.com/novodex_downloads.html <- click!

is there really a need for specialized hardware like this ppu in times of multicore cpu's? i mean, let intel or amd simply add another core which will take the processing part of the ppu and be done with it.

This calls for a more realistic version of Porrasturvat \o/

the mind reels at the thought of what this might do for the presswurst series.

The reason why PPUs won't have similar huge impact as GPUs is very simple: every game has pretty much the same graphical operations (rendering triangles/quads with texture and lighting), whereas physics algorithms can vary hugely from game to game.

I'd compare a PPU to a GPU that supports scene graphs: sounds neat at first, but in practice there has to be room for game-specific tweaks.

Of course, a good PPU design might be general enough, and future games (at least mainstream 3d ones) may have very similar physics algorithms, thus opening the market for PPUs.

biggest problem is: how to support both kinds of users (with/without PPU card)? cause it's by no means as easy as changing the rendering options for a slow GPU (f.e. disable shadows, etc. for a speed boost)!

"The reason why PPUs won't have similar huge impact as GPUs is very simple: every game has pretty much the same graphical operations (rendering triangles/quads with texture and lighting), whereas physics algorithms can vary hugely from game to game."

Thats what most people said about the first GPUs aswell.