Graphic Cards nVIDIA PhysX vs DX11 physics

The CUDA is an open spec. Anyone can come up with their own implementation of it.. So, if ATI really wanted to support GPU accelerated Physx, they just need to write a CUDA driver, but obvious market politics means that they wont :-|.

An up and coming API that I feel has a good future is the DMM engine from Pixelux. In fact they are currently working on an OpenCL implementation of the same which they intend to integrate with Bullet from ATI.
 
stalker said:
The CUDA is an open spec. Anyone can come up with their own implementation of it.. So, if ATI really wanted to support GPU accelerated Physx, they just need to write a CUDA driver, but obvious market politics means that they wont :-|.

An up and coming API that I feel has a good future is the DMM engine from Pixelux. In fact they are currently working on an OpenCL implementation of the same which they intend to integrate with Bullet from ATI.
Click to expand...

Market politics is not the reason, they will be losing market share if a lot of PHYSX accelerated games are released. It may be much tougher to implement PHYSX on their hardware me-thinks, it just does not make sense otherwise. PHYSX will surely be built upon NVDA hardware, and since the past few generation the differences between AMD and NVDA hardware has widened significantly.
 
Aces170 said:
Market politics is not the reason, they will be losing market share if a lot of PHYSX accelerated games are released. It may be much tougher to implement PHYSX on their hardware me-thinks, it just does not make sense otherwise. PHYSX will surely be built upon NVDA hardware, and since the past few generation the differences between AMD and NVDA hardware has widened significantly.
Click to expand...

Physx can run on any chip that can run CUDA. It is not tied down to the chip it runs on. Also, like I previously said, you don't need a nVidia GPU to do CUDA, its an open spec.
 
^^ Yep get that part, but the coding the AMD hardware would be more complex in CUDA, else there would be no reason for AMD not to implement it, its a win-win game for them...
 
I would think the basic reason ATI is not using Aegia PhysX or ever will : "why use a competitors technology". I am sure, they have the expertise to enable their cards to understand the engine.
 
An interesting quote that I found over at Charlie's forums in a discussion of the same topic

You are going to program CUDA on ATI GPUs, and would need to download a CUDA driver from NVIDIA's site, and you saw this message at the download page:

Quote:

For the best performance and all of the latest CUDA 4.0 features, the latest NVIDIA GeForce GTX 400 series GPU is required, click here to check out the latest lineup of NVIDIA GTX 400 series GPU.

Currently Radeon series GPUs can only support up to CUDA-ATI 3.1.
Click to expand...

Which to your knowledge these two variants have the same feature set for the same version number, only the optimization is different.
Click to expand...

The thread itself -> CUDA on ATI? - SemiAccurate Forums
 
Aces170 said:
^^ Yep get that part, but the coding the AMD hardware would be more complex in CUDA, else there would be no reason for AMD not to implement it, its a win-win game for them...
Click to expand...

I second that.. STREAM is not similar to CUDA, it would be very tough for developer to implement PhysX on Stream and they have to write two peices of codes to get it to work..

Moreover, AMD has always slamed PhysX as Proprietary strandard rather than being an open.. It may be they are using word "Proprietary" in that its not easily programmable or may be they come with the open standard as excuse and lies..:)
 
muzux2 said:
I second that.. STREAM is not similar to CUDA, it would be very tough for developer to implement PhysX on Stream and they have to write two peices of codes to get it work..

Moreover, AMD has always slamed PhysX as Proprietary strandard rather than being an open.. It may be they are using word "Proprietary" in that its not easily programmable or may be they come with the open standard excuse and lies..:)
Click to expand...

Saying that something is tough to code is nothing but an excuse. Also, there is no reason for AMD to implement PhysX using Stream.

Secondly, the thing with CUDA is that it is an Open spec in that it is freely available to anyone, however, it is not maintained by an impartial 3rd party. nVidia clearly has a vested interest in it, this means AMD will never be party to using/implementing it.
 
asingh said:
I would think the basic reason ATI is not using Aegia PhysX or ever will : "why use a competitors technology". I am sure, they have the expertise to enable their cards to understand the engine.
Click to expand...

well, that can be the reason..just imagine an ATI graphics card box with an Nvidia CUDA or PhysX logo on it...:lol:
 
erm.. accepting that the competitors product is better is win-win for AMD?
Click to expand...

Not a product, but a platform. They are getting a development tool free of cost to improve their product, will hardly lead to ego issues, especially with the current cash crunch they are facing.

Saying that something is tough to code is nothing but an excuse.
Click to expand...

Not if it involves a lot of resources.
 
AMD Developer Forums Forums - OpenCL performance issues

We're developing using openCL, and have one dev machine with an NVIDIA GTX 260, and another with an ATI 4870. These both seem to be mid-range cards, and are similarly priced.

However, I'm sorry to say we are getting approximately 5x the performance from the NVIDIA card, than from the ATI. We're using the same openCL kernel, and the SDKs of the respective companies - in the case of the ATI, Stream SDK 2.0 beta 4.

Is this performance gulf due to the early stage of ATI's OpenCL support? Is the implementation not well optimised yet? If so, how soon can we expect this gap to close? Or were we mistaken about the equivalence of the two cards?

To be honest, I was expecting more from the 4870 - certainly the specs seemed to imply it was fairly powerful - however, currently my CPU (core2 quad 2.4GHz) can outperform it by roughly a factor of two. Something definitely seems amiss!

Many thanks for any information
Click to expand...

Food for thought....
 
iGo said:
@Raghu:

@atiamd: I don't think looking better has anything to do with physics engine being used. Physics engines such as Physx or Havok add more realism to game by making environments behave as close as real life. Stuff bouncing off the surfaces like wall or floor like it would do in real life or shooting the enemy and watching him fall like real person are kind of stuff physics engine is supposed to take care of.

The key difference is, who or what does all this physics processing. Running it off the CPU isn't the best way as it creates significant impact on gameplay performance. GPUs on the other hand, have massive number of parallel processors which can run this task with minimal impact on framerates. At least that's the basic idea behind running GPU bound physics processing engine. Having said that, a well coded game in Havok is going to look almost as good as Physx, but while Havok will strain your CPU as complexity in physics processing increases, Physx will be able to keep performance hit minimal as it will use your GPU for the job.
Click to expand...

Cant believe you just said that. After reading my posts in teh current thread, you really think that I dont KNOW what physics does in games?... daaamn. :hap5: I appreciate the time you put in typing though...
What I really meant to ask was if physics from physX does in fact look much better than havok or anything else or not? I asked this coz I have seen all the other engines mentioned in action, but physX. Hence my question...

Anyways, I guess I have nothing new to add to this thread. So, whenever I get something new, I will report back........:hap2:
 
stalker said:
Physx can run on any chip that can run CUDA. It is not tied down to the chip it runs on. Also, like I previously said, you don't need a nVidia GPU to do CUDA, its an open spec.
Click to expand...

but PhysX is not and there ends the story of Physx on ati.It can run on any chip that can do CUDA does not mean that it will be allowed to.It's just like the AA in batman, only with more legal/ethical backing.

Aces170 said:
AMD Developer Forums Forums - OpenCL performance issues
Food for thought....
Click to expand...

48xx series opencl support is still in beta, and will never be as good as the gtx2xx series considering that the hardware was not developed with the opencl spec in mind unless you go around the hardware shortcomings, which, considering ATI's support on the subject, will not be an inviting prospect for the developers.
 
y wud MS being a player in the field of software right a code for PS3 and wii both of which run their own propreitory software in their respective consoles .

Dont think sony and intendo wud appreciate MS wiriting code for their consoles .DX11 physics is written for its own OS period it does not have to be cross platform they just waNT game physics to not be run only on nvidia hardware .

Keep in mind MS is also in the console business ur arguement makes no sense whatsoever try to think before u start arguin dont do it for the sake of it !! belittling others wont make u any better ...
 
Back
Top