pouët.net

and.. yes, if they manage to bulk this with some heavy game from idsoftware/valve we are prety much stuck to its standards for some years anyways.

Poop Producing Units already have much impact on the scene.

the concept is just very nice. i think that when they'll fit it in a single chip it will be much easier to be spread... and on other platforms if could be emulated by sw, like opengl in its first incarnations did.. anyway, i think that something like this can have a nice future - of course, the API flexibility will matter a lot imho

this means that FAG can soon do realistic bukkake demos.

The PPU costs between $200 to $300? I don't bloody think so. Just buy a NES console off eBay for a few bucks and pry the PPU from the PCB. Voila!

man need a new 3Dfx :)

http://www.firingsquad.com/hardware/ati_physics/

jsyk.

using the gpu as a big, fast, parallel fpu isnt really new.. but it's only just becoming really useful now you can e.g. render to vertexbuffer and lookup textures in vertexshaders, so you dont have to copy it all back to cpu in between.
will be nice when that hardware makes it to being the "standard", and we can all really start abusing it.

up.

I wonder if this could help softsynths to sound less like Virus and Juno and more like decent physical modelling.

about demos using this card, it would surprise me to see demos using hardware accelerated physics... my (a bit on the fly meditated) argument: when democoders started using 3d acceleration they already had a long background on 3d software rendering, they were familiar with the principle and most importantly 3d graphics were a key stone on a demo. I don't see similar process with physics. There is almost no demo with (software) physics yet (ok, there are a very few), even if there is many libraries available (ODE, Newton, whatever). So, physics seem not to be relevant for demos. So, I don't see why a HW version should change that.

You have that creative use of the HW to do different thinks, but still depends how programmable the card is. If the API is just "xxNewWorld, xxAddMesh, xxAddCube, xxCollide, xxStep", then I don't see much options for creativity, or no more that what you have with ODE since 5 years or so.

For HW raytracing, I would bet for ATI implementing some sort of local-to-the-mesh raycasting functions in GLSL set (for small meshes first of course)... That would be a step forward in CG.

Also in the physx case you need the user to install the ageia drivers, even for the software version -> instant thumbdown galore. And we know it's all about getting as many thumbups as possible.

after the GPU... the NKVD

:P

heh ithaqua i remember the game GRAW that needed to install phsyx drivers even if i had no ageia card, i was like wtf! that's compelled marketing :/

I'm thinking I could enjoy a modern version of The Incredible Machine... :)

http://www.tokamakphysics.com/ is a nice physics engine.

I propose a PHIGS+ / IrisGL style API war over near-identical physics simulator implementations, resulting in the loss of the most useful high-level concepts. Ready... set...

3DFX is dead.

Now:

Winner = Speed ( Code[what works everywhere] )

HURRAY!!