Sony adopts Ageia's physics chip for PS3

Archived from groups: comp.sys.ibm.pc.games.action (More info?)

Story at tomshardware
http://www.tomshardware.com/hardne [...] 40938.html

Wonder if this deal could push down the price of the PC add in cards?
Probably not but what the hey?

Archived from groups: comp.sys.ibm.pc.games.action (More info?)

On Thu, 21 Jul 2005 16:10:54 -0400, "Magnulus"
<magnulus@bellsouth.net> wrote:

> It's not clear exactly if Sony gets the chip or just the SDK- I'd
>imagine the Ageia physics stuff could run on one or several of the DSP's the
>cell processor has. Though the article makes it sound like it will have a
>chip integrated.
>

The article is written by somebody who could not comprehend the
announcement and/or made a leap of wishful thinking.

See:-

http://ageia.vnewscenter.com/conte [...] 0108406252

( You may need to unwrap the above line...)

Just the multithread SDK for use with the Cell processor, not the
chip. Want a PS3 to cost at least $600 ?

Sony also licensed the multithread Havok SDK.

Seems as if Sony is corralling the industry's best game-development
tools for themselves.

Wonder what the Xbox360 has? Oh, silly me... M$$ development tools,
of course......... :-) :-)

> What I want to know is why it takes a small 3rd party company to get this
>stuff going?

Yes indeed, Havok.... (BTW, founded by a couple of professors from
Trinity College, Dublin, Ireland.)

Now Ageia and others.

John Lewis

>You would think Microsoft or the open-source community
>(OpenGL) etc would have had the forsight to develope a
>non-hardware-proprietary physics API. It's bad enough that Microsoft
>dropped the ball with DirectX and DirectSound, and basicly let Creative be
>the only player in town- now it's really too late to do anything about it.
>
>

Archived from groups: comp.sys.ibm.pc.games.action (More info?)

Magnulus wrote:
> The only thing that concerns me about the Ageia PhysX chip/card is that
> initially they are going with a 13nm process with 110 million transistors,
> and it will have a relatively large heatsink/fan combo, about the size of a
> small graphics card, requiring a molex connector. It seems to me they
> should spend the extra money and get a 11 nm process and maybe paired down
> the initial performance of the first card. That would allow a slim solution
> with possibly only a small heatspreader/heatsink to fit in a PCI slot, and
> possibly without a molex connector. Initially, all the Ageia card is going
> to do is allow extra performance in games, it simply isn't going to have its
> full potential right away, in the same way that 3D graphics cards did not
> immediately increase the resolution games could be played at, because
> developers can't specificly code for it alone.

You are a factor of 10 off the gate dimensions. It is 130 nanometer,
110 nanometer and 90 nanometer. The numbers you are using are in the
far distant future and may actually never happen (the barrier to the so
called Moore's Law arriving long before those dimensions are reached).
90 nanometer is the state of the art right now and does result in less
energy dissipation but the wafer yields are kind of iffy right now.
Plus there is only one fab in the world that is achieving acceptable
levels of wafer yields for 90nm. The question is not if it can be done
but rather how much does it cost to do it. That is what is going on
between Nvidia and ATI at the moment. The Geforce 7800GX is 110 nm,
resulting in almost 30% power requirement reduction compared to the
6800 Ultra at almost twice the performance. ATI decided they wanted a
90nm process for the R520 but that has come at a serious cost in time
to market. Nvidia is working on a 90nm process so they may be able to
outstage ATI when they finally come to market. Ageia is pretty much a
startup, so using 130nm which was state of the art last year is
understandable. It is about affordability while working into market
acceptance.

Archived from groups: comp.sys.ibm.pc.games.action (More info?)

Oops, sorry about the decimal error. It should be 0.09. And NO, I don't
work for NASA.