ATI's XFire/Physix card QUESTION's

One idea is having three card, one for physics, and two for rendering graphics. The other idea is having two card, one for physics, and one for rendering graphycs.
This article is nice, and should explain most things.

Right now the PPU is pointless. It does nothing. Future games will surely take advantage of a PPU, but the video cards will take care of that. Right now, there are no benefits from having the PPU, but that should change over time,

thx for the info so if anything i will be geting a 1900xt or ill just wait that 6 months or so for the newer dx10 ati's gpus.....btw does anyone know if the prices of those gpus will be high? or they will be similarly priced as the 1800 and 1900 series were prices upon their release?

thx

WE can't tell the prices, but I predict they would be around $400-$500US for top-line cards. About the same as X1900/X1800 and 7900GTX.

all cards come out within those price range...

it mainly ranges from 400$-700$ for top cards

I strongly dislike ATI's physics solution, for it seems a bit hypocritical to me; they criticize AGEIA and then want a dedicated GPU for physics calculations? From what I've been hearing from the Nvidia camp, they'll be using the unused pixel processing power for their physics solution, saving me money on my next purchase.

I agree with prozac basically it's so they can sell more cards right guy's? I mean look at it 3 x1900's for about 400 a peice who the heck is gonna pay for that?(Maybe some die-hard gamer but not me)

From the nVidia camp?
Right now nV's not even promoting unbalanced GPUs!

And considering SLi phsyics is pretty much dead as it's own implementation then the limits/benifits of Havok's FX will apply to both companies.

And consideirng the design of the R600, I'd expect them to have more flexability in the immediate future than nV.

But after all of the AGEIA bashing, we're now going to buy an extra videocard to do physics calculations? As I said, it seems a bit hypocritical.

Except you don't need to buy a $300 card like the Ageia card, the $49 X1300 would do a better job according to ATi's (probably biased) numbers. And while I agree that I would prefer using the unused power, or even integrated, it's unlikely that ATi is going to promote that as the best possible solution, when they can also get more money from more cards. However if it's available elsewhere it'll be available from them as well. Which is why I say nV will bring unbalanced options just like ATi, not because they've announced them themseleves, but because Havok supports them, and it's unlikely they will let that opportunity go by. But don't forget that they too pushed the SLi aspect strongly.

The question is FLEXABILITY, and when you think about it, it makes more sense to have it be a somewhat balanced solution so you can get the Xfire benifit, like an X1900GT + X1900CF card would let you use the GT for physics, and then for games that don't use physics they could run in Xfire mode. And that would be for less money than the PhysX card and it ould be more global benifit in many more situations. Still limits (likely oly felt in flying and racing games IMO) but I doubt those limits would be of great concern to most people and are nowhere near the limits of the PPU.

All valid points. But how does quad core CPU's and beyond factor into this debate?

My belief, however good ATi/Nv is at providing Physx support by means of 2nd/3rd card, is that the GPU does only graphics rendering, and the CPU takes care of the rest. What's stopping intel/amd from creating a quad and above cpu that has 2 core's doing just physx and leaving sound to a dedicated sound card and graphics to the already good gcards from both camps?

Sure, the gpu excels at simple/focused tasks, but what's stopping either cpu camps from giving good code for physx forward?

Just a thought - I think everyone should stop (Ageia/Intel/Amd/Ati/Nv) and reach some sort of common ground and decide who's doing what - cause right now everyone wants a piece of physx pie, and the consumer is left with a difficult choice of what to buy to have a good running gaming rig. BUT, the chances of common ground being reached is like a vw bettle winning the le mans....

Firstly, I want to add something on the CPU front. I don't think keeping physics running on CPUs is a good idea... UNLESS the next-gen CPUs get a huge boost in floating point calculation power, which is what is needed for dynamic physics processing. Havok is a good solution, but because it relies on CPU power, it is limited in what it can do. Even a Quad-core processor (unless it gets more floating point processing power) will struggle when exposed to what Ageia claims that their PPU can do.

Personally, I think ATI's solution is great... for the moment that is. At least I know I can keep my beloved X1900XT and use it along side a DX10 ATI card when they come out (Hopefully I can do that), letting my X1900XT doing all the physics calculations, and the DX10 card dealing with the graphics. However, by then, Ageia could have gained a bigger development base and pushed more solutions, and gained more support with games, so whatever happens, its gonna be tough for the consumer. The only hope is for a DirectX-equivalent standard to allow compatibility between solutions (Another hope).

To me it's simple. The CPU will always be best at game dependant physics if it can handle it (one again driving and fling games might need help), where the truely interactive bits would be best on one of the 3rd or 4th cores (maybe sharing a mapping on networking chore). The GPU or PPU could still do all the massive increase in particle shiny physics. which are really the whipped topping everyone seems to be impressed by (Realistic looking explosion, realistic looking flames, etc. the focus usually being looking).
I think the level of game dependant physics is usually minor in comparison, something that multi core or additional FPUs (ala CELL) could easily handle without breaking a sweat.