Graphic Cards PhysX to kick ATI in the guts HARD!

EscSysRq

EscSysRq

Adept
Talk is the ATI were offered PhysX but declined because they thought Havok was a better implementation.
Could they have been wrong.......Cryostasis, Mirrors Edge...just to name 2..

What do you guys think, did ATI blow it big time and will their decision to not licence PhysX bite them in the A$$ hard.
Personally I think they may have blown it on this one, I will certainly look at nVidia next gen and PhysX is a big factor in that decision.

I pick this from here ...

PhysX to kick ATI in the guts HARD! - Rage3D Discussion Area ?
 
EscSysRq said:
Talk is the ATI were offered PhysX but declined because they thought Havok was a better implementation.
Could they have been wrong.......Cryostasis, Mirrors Edge...just to name 2..

What do you guys think, did ATI blow it big time and will their decision to not licence PhysX bite them in the A$$ hard.
Personally I think they may have blown it on this one, I will certainly look at nVidia next gen and PhysX is a big factor in that decision.
Click to expand...

Source? :)
 
Could they have been wrong.......Cryostasis, Mirrors Edge...just to name 2..
Click to expand...

Not to side with ATi or something... but those two are developed in association with nVidia if I'm not wrong. Havok or Physx, from our, consumer's standpoint we are not fully benefited until there is vendor agnostic implementation.

It's the same thing as CUDA on NV or Stream on ATi... thankfully, OpenCL and DirectX 11's compute shader will enable GPGPU computing on either vendor's offering. Unless there is something similar comes along the overall physics computing scenario will be pretty much limited. If you've seen Mirror's Edge videos demoing physx effect, you will notice that there are too many things missing in non-physx video. I mean forget physx interactions, entire object (like drapes, plastic covers on the edges) are gone if physx is not enabled. That's some serious cutdown from game environment to remove the entire objects in non-physx play.

So in my honest opinion, software makers should concentrate more on universal or vendor agnostic technologies more than proprietary technologies.
 
^^ agree with you..

To adopt PhysX, ATI has to first adopt CUDA. i highly doubt ATI will never adopt it coz CUDA isn't open. ATI has its own GPGPU computing known as STREAM. Accepting CUDA would be an insult to their GPGPU program.
Just imagine an ATI Card box with an Nvidia CUDA/PhysX logo on it. :rofl: It's never going to happen.

If PhysX could be ported to openCL, then i believe ATI can support it, since both of them supports OCL. but first, Nvidia has to make CUDA open.

To me, PhysX brings only fancy effects to game & doesn't have an impact on the way we play the games..:)
 
^That too with stunning physics processing. Mercenaries 2, John Woo's Stranglehold, Saints Row 2, Fallout 3 etc. just to name a few. Stranglehold actually showed excellent physics demonstration with literally everything being destructible.
 
iGo said:
If you've seen Mirror's Edge videos demoing physx effect, you will notice that there are too many things missing in non-physx video. I mean forget physx interactions, entire object (like drapes, plastic covers on the edges) are gone if physx is not enabled. That's some serious cutdown from game environment to remove the entire objects in non-physx play.
Click to expand...

In mirror's Edge ,when PhysX is turned off obviously the Physics computing objects will not be visible.Well example for that will be, If we disable the Havoc computing in games then what will we expect. :)

muzux2 said:
Just imagine an ATI Card box with an Nvidia CUDA/PhysX logo on it. :rofl:
Click to expand...

:rofl: All is possible for ATI.. Remember when AMD took over the ATI. At that time i think you might be saying these words "It's never going to happen." :p
 
muzux2 said:
^^ agree with you..

To adopt PhysX, ATI has to first adopt CUDA. i highly doubt ATI will never adopt it coz CUDA isn't open.
Click to expand...

You have no idea what you are talking about. The spec is open and the compiler is also open source.
 
CUDA is a programming language if I am not mistaken, what do you mean by when you say its not open, you mean to say Intel pay's C developers everytime they ship a proccy ?

The only reason ATI is not doing is obviously CUDA is built around to exploit NVDA's architecture not ATI. But since GPU's have become processing powerhouses we will need a universal programming platform which I think DX 11 will address...
 
Raghunandan said:
You have no idea what you are talking about. The spec is open and the compiler is also open source.
Click to expand...
Aces170 said:
CUDA is a programming language if I am not mistaken, what do you mean by when you say its not open, you mean to say Intel pay's C developers everytime they ship a proccy ?

The only reason ATI is not doing is obviously CUDA is built around to exploit NVDA's architecture not ATI. But since GPU's have become processing powerhouses we will need a universal programming platform which I think DX 11 will address...
Click to expand...

CUDA is open in the sense that only its specification can be downloaded but it's not open coz no other hardware vendor supports it or contributes to its development. Don't get me wrong, even Godfrey Chang(Director of technical marketting, ATI) said that PhysX will die if it remains a closed & propreitary standard.. Had PhysX been open, then why is ATI saying it all..

AMD: PhysX Will Die, Havok Is the Future - The company stresses on a better integration of GPU physics into gameplay - Softpedia

techPowerUp! News :: PhysX will Die, Says AMD

:)
 
^^ But then that's what they will say... coz they are not supporting it and opting to go with their own or rather stick to what intel would do. :)

I mean I haven't seen rivals applauding others technology before.. have you??

As for NV... I really like them for their stupidest yet bravest decisions in past.
 
Not to mention that they are not exactly allowing this tradition of thiers to die a quick death... What with the new shutter glasses nd all.. :mad:
 
OpenCL, on the other hand, offers developers the ability to write an application once, compile it once, and expect it to run on all major GPU hardware. Something that could never happen with ether CUDA or Brook+.
Click to expand...

this is what i was telling here, if both brook+ & CUDA can be ported to OCL, then we can have one standard for physics acceleration.. Does that mean, PhysX will die when DX 11, Compute shaders with OCL comes around..guys, don't get me wrong, I am not slamming CUDA here...

isn't Mr. Cheng then right, he says
As we have emphasised with our support for OpenCL and DX11, closed and proprietary standards will die.
Click to expand...
..:)

it seems then, PhysX will succeed in the short term (the way its going presently) and looks dubious about the long term..

i think, ATI is blowing smoke, buying time until DX11 when compute shaders become standard and they will get free access to GPU-accelerated physics..:rofl:
 
No Physics API will be THE standard just like a game engine. Every game has its own requirements, some games would do expensive stuff like cloth and fluid particles but many games won't so it really boils down to the game developers what they want to use.

I think compute shaders(DX11) is going to be a standard for GPGPU related to games. Physics API like Havok & PhysX will be ported to this API and everyone will be happy.

Btw, ATI has already started working on OpenCL, but no one knows whether it(OpenCL) will be popular, as we have seen Direct3D getting more popular against OpenGL, but then again ATI cannot take any chances.
 
Forrest said:
but no one knows whether it(OpenCL) will be popular, as we have seen Direct3D getting more popular against OpenGL
Click to expand...

I think you're confusing OpenCL with DirectX and OpenGL. OpenCL is not graphic subsystem API. It's Open Computing Language, which let developers easily tap into vast resources of graphic card to GPGPU computing, while totally being vendor agnostic. So you're not limited by proprietary solutions offered by different hardware vendors tied to their own products only.

For eg. any app created on CUDA will run only with GeForce cards and won't work on others like ATi or Intel graphics. Same goes for ATi's Stream. Intel had proposed very promising solution that was Larrabee and X86 programming interface, but since no one has seen Larrabee yet no one knows how it's implemented. OpenCL will work with all these vendors products, via common driver interface so you're not limited to using particular hardware for certain app. Like, you can't play mirror's edge with all physx interaction enabled on ATi graphic card.
 
So you're not limited by proprietary solutions offered by different hardware vendors tied to their own products only
Click to expand...

Thats what they said about OpenGL too when it came out and now it bites dust compared to Direct3D. It really doesn't matter whether the API is open or not, if its not as good as compute shader then it won't be popular.

Oh and I am only talking about games here as OpenGL is still widely used in visualization applications.

OpenCL will work with all these vendors products, via common driver interface
Click to expand...

You have no idea what you are talking here. There is no comman driver interface, OpenCL API has a set of functions and each vendor has to write those functions in their own low-level language, for e.g, CAL for ATI, its basically like a compiler which changes a source language to a target language, so OCL will be changed to CAL and then compiled.
 
Back
Top