AMD Announces Open Physics Initiative

[NickK]

Hehe I missed this thread

This will only add additional pressure to nVidia's market penetration. They really need to rethink their approach and twiddle a few strategic knobs otherwise they'll have cashflow issues as developers will plum for the open/MS technologies.

So all the games manufacturers love DX11, not surprising for the xbox.. this puts AMD in the driving seat for the next generation of xbox sales.

I've no doubt that DX will become very popular and that OpenCL will too (just as OpenGL has).

 

[tdream]

AMD: NVIDIA PhysX Will Be Irrelevant

AMD said they do not consider the lack of PhysX on ATI Radeon graphics card a problem, the company believes the PhysX technology will become irrelevant in the future:

Advanced Micro Devices, the world’s second largest developer of x86 central processing units (CPUs) and a leading designer of graphics processing units, said that it hardly regrets about Nvidia Corp.’s decision to disable support of hardware physics effects processing using PhysX API and GeForce GPU or Ageia PhysX PPU in systems where ATI Radeon graphics card is used for graphics rendering. In fact, AMD believes that with the raise of popularity of DirectCompute and OpenCL APIs, proprietary PhysX will soon vanish into oblivion.

“Physics can be a good thing for gamers, but it should be for all gamers. When it’s available for everyone, game developers will be able to make physics an integral part of gameplay, rather than just extra eye candy. This requires a physics solution built on industry standards. That’s why DirectX 11 is such a great inflection point for our industry – DirectCompute allows game physics that can be enjoyed by everyone. There are several initiatives (some open-source) that will deliver awesome GPU-based physics for everyone, using either DirectCompute or OpenCL. Industry standards will make any proprietary standard irrelevant,” said Neal Robison, director of global independent software vendors relationship for AMD.

“There’s a real discrepancy between what Nvidia says, and what they do. They “say” that they are looking out for gamers’ best interests. However, decisions like this are the exact opposite of gamers’ best interests,” added Mr. Robison

http://www.guru3d.com/news/amd-nvidia-physx-will-be-irrelevant/


Also agree with Rroff's previous post about physics will be a driving force in gaming. Especially in pc gaming, that's where you can seperate between console and say yes, that's clearly much better than the same game on a console.

 

[kylew]

AMD: NVIDIA PhysX Will Be Irrelevant


http://www.guru3d.com/news/amd-nvidia-physx-will-be-irrelevant/


Also agree with Rroff's previous post about physics will be a driving force in gaming. Especially in pc gaming, that's where you can seperate between console and say yes, that's clearly much better than the same game on a console.

Unfortunately, that will never be the case.

For physics to have a big impact on the way games work, they need to be implemented at the roots of the game basically.

This is why PhysX has been so poor, because it's always been an add-on effect, they can't make games that require PhysX as minimum because not everyone could play it.

Once the majority of people's GPUs can accelerate physics on a GPU, then we'll start seeing game changing physics implemented.

That means that consoles will need to be able to run physics code on the GPU too, otherwise multiplatform games are always going to be restricted by the weakest hardware.

 

[Rroff]

Unfortunatly kylew is right - most developers these days concentrate on developing for the console coz thats the cash cow - they aren't going to develop a game that has physics integrated at every level unless it will run on consoles like that too... so any physics effects on the PC in those games will be incidental... tho those games developed natively for the PC platform that do implement widespread physics will probably leave the console games looking decidedly dated.

I don't see this move by ATI necessarily making physX redundant - otherwise havok would have done that a long time ago (stifled it before any proper hardware accel. back in the novodex days) - bullet currently lacks the features and implementation style that works ideally for a lot of games (its original focus I believe was more for movie effects) ... but... this move could result in it becoming a focus of attention for implementing and refining those kinda features.


EDIT: That said I know theres been a lot of interest and requests for hardware physics acceleration from many studios so I wouldn't be suprised if next gen consoles do include such functionality.

 

[tdream]

Unfortunately, that will never be the case.

For physics to have a big impact on the way games work, they need to be implemented at the roots of the game basically.

This is why PhysX has been so poor, because it's always been an add-on effect, they can't make games that require PhysX as minimum because not everyone could play it.

Once the majority of people's GPUs can accelerate physics on a GPU, then we'll start seeing game changing physics implemented.

That means that consoles will need to be able to run physics code on the GPU too, otherwise multiplatform games are always going to be restricted by the weakest hardware.

Never? Never is a long time. Come back in 10 years time and see how much physics there are in games.

I agree with you about Physx but then I already knew that.

Unfortunatly kylew is right - most developers these days concentrate on developing for the console coz thats the cash cow - they aren't going to develop a game that has physics integrated at every level unless it will run on consoles like that too... so any physics effects on the PC in those games will be incidental... tho those games developed natively for the PC platform that do implement widespread physics will probably leave the console games looking decidedly dated.

I don't see this move by ATI necessarily making physX redundant - otherwise havok would have done that a long time ago (stifled it before any proper hardware accel. back in the novodex days) - bullet currently lacks the features and implementation style that works ideally for a lot of games (its original focus I believe was more for movie effects) ... but... this move could result in it becoming a focus of attention for implementing and refining those kinda features.


EDIT: That said I know theres been a lot of interest and requests for hardware physics acceleration from many studios so I wouldn't be suprised if next gen consoles do include such functionality.

Of course the shape of the market today is like that, you were talking about the future right? Odd that I'm trying to validate a point you made...

Ati isn't making physx redundant, physx itself is unless it goes open source, which opencl and directcompute are. If you have a choice between xbox/ps3 batman and pc batman and pc batman has open souce physics that any gpu can run, the pc option becomes a lot more attractive. Physx is physics so the incidental examples we're seeing are just the beginning but they do add to the game. They bring more atmosphere and can easily be used in puzzle solving segments or games like portal where physics can easily slot in.

Re: your edit, which was your point in the first place.

 

[kylew]

Never? Never is a long time. Come back in 10 years time and see how much physics there are in games.

I agree with you about Physx but then I already knew that.



Of course the shape of the market today is like that, you were talking about the future right? Odd that I'm trying to validate a point you made...

Ati isn't making physx redundant, physx itself is unless it goes open source, which opencl and directcompute are. If you have a choice between xbox/ps3 batman and pc batman and pc batman has open souce physics that any gpu can run, the pc option becomes a lot more attractive. Physx is physics so the incidental examples we're seeing are just the beginning but they do add to the game. They bring more atmosphere and can easily be used in puzzle solving segments or games like portal where physics can easily slot in.

Re: your edit, which was your point in the first place.

Oh no, sorry, I didn't mean never ever, I meant never with regards to this current generation of consoles, and never until they have the hardware to support hardware physics.

There'll be a point where it's in their best interests to add hardware physics support to consoles, until then, it's never gonna be a reality that multiplatform games have game changing physics implemented on the PC only given the already lacking hardware of the consoles.

 

[Rroff]

Sorry think I mis-read some of the points presented in posts above... and skipped over a few waymarks myself so as not to get sidetracked into arbitary arguements... and now I'm tired, drunk and prolly not making any sense at all :S

As kylew said the same game on PC and Console isn't going to be remarkably different as until there is physics hardware on consoles they can't do anything majorly game changing in a title even on the PC port unless they spend a huge amount of time rewriting it for PC which I don't see happening...

However once there is a killer title for the PC tho that uses this stuff I see it making console titles look decidedly dated and will probably open the flood gates, developers will clamour for the support on consoles and we should see a major shift.

nVidia has pretty much killed off physx as such with their recent moves... but despite that I can't actually see it being made redundant... theres a reason that physx has a relatively large following in that its often the most suited physics engine to actual games without being intrusive or distracting - havok while technically very good and does have a lot of developer following tends to get in the way a bit and/or not do things in a way that works as well in video games as it would in real life and bullet as I said seems to be more focused on cinematic style implementations that again don't translate as well to gaming scenarios... not to mention its not as mature or polished as either of the other 2.

 

[cyberheater]

I have a dream that one day I will play a fps and no one falls thru a wall or clips through an object ever... Once they sort that out, then they can add the nice stuff.

 

[NickK]

nVidia's control in the games market is limited by MS and xbox.

I see games developers adopting this when MS product a physics engine in DX. The runtime is then up to windows.

I see the AMD open physics engine also separating the application developer and the runtime, leaving it to the operating system.

Interesting that Apple will probably go open for the physics engine initially too by using the OpenCL platform.

The next generation of operating system GUIs will not just be using graphics they'll be using physics to provide a natural organic feel in use. It's use is not just applicable to games..

 

[Axil]

I can only really see physics in games only going fully mainstream when the consoles have it as part of their SDK with hopefully a open source implementation to enable the cost of putting it into a game to be reduced.

PhysX to me is a little bit like the situation we had when 3D GPU games were starting out with every graphics company producing their on SDK which meant that usually the only game really supporting them were the games actually coming with the card itself.

For me no developer in their right mind would implement a physics solution that only covers half the graphics cards out there as it will just limit their possible sales (unless of course said physics SDK supplier pays for the shortfall caused by the lost in sales)

 

[Shakeel]

good for competition

 

[Duff-Man]

Yeah, I agree with the points about consoles needing to support GPU physics before it will become truly widespread in games. Unfortunately the state of the console market has a big impact on the PC these days. I think an open physics standard will allow the PC to provide a "proof of concept" though, and encourage full support for hardware physics in the next generation of consoles. Once the next-gen console specs are defined and announced, developers will have the confidence to start trying to exploit the extra hardware capabilities on the PC.

On a side note... Jesus, have you ever seen so much general agreement in a thread on this forum?!

 

[kylew]

Yeah, I agree with the points about consoles needing to support GPU physics before it will become truly widespread in games. Unfortunately the state of the console market has a big impact on the PC these days. I think an open physics standard will allow the PC to provide a "proof of concept" though, and encourage full support for hardware physics in the next generation of consoles. Once the next-gen console specs are defined and announced, developers will have the confidence to start trying to exploit the extra hardware capabilities on the PC.

On a side note... Jesus, have you ever seen so much general agreement in a thread on this forum?!

Most definitely not, especially between my self and Rroff.

 

[NickK]

AMD stream computing, and now OpenCL, has always been pushed at the film effects production industry. It's a big customer with deep pockets.

The revenue driver, the package of film and console game, is more and more important. The game is today's action figure of yesterdecade (think starwars!). Future "media" will be actual interactive cinema where the film and the game share the same virtual platform.

Apple have pushed into that area and with Jobs now owning pixar and a nice slice of Disney (I suspect he has his sights on CEO there too), it's Apples push into the effects business that AMD would love to get their fingers in.
AMD have attempted to play a Cinema "me too" play to Apple to push out Intel and to ensure Intel don't capture the xbox revenue stream.

I am sure that Bullet will evolve into a solution, however it's a leading prototype to show that they can deliver physics for the Xbox/Apple etc and gain traction. Their big player partners - Apple and Microsoft - will product a standard interface for their developers.

The more I see of AMD, the more I think their business strategy is purely "me too" where they play the number two in the market - even when their products blitz it (opteron, some of the ATI stuff). This strategy is because they want their partners to be number one with their IPR and width of markets to provide their revenue..
Only thing is it doesn't give that "cosy" feeling of someone owning and driving vision that shape the products like Apple, MS or Intel have.. nVidia have started 'owning' their space with their new set of tools but they don't have the market to sell to developer to balance the business case.. they don't own an operating system platform.

Hmm.. I think I may re-read my Porter, Moore and Johnson/Scholes books on strategy.. been too long since I've had fun in this area

 

[unknownsock]

Someone wanna correct me here plx.

So ATI cards are already physics capable but only in OpenCL?

WHat about old PhysX games? They will have to be ported to OpenCL for ATI cards to ever really work with this?

Dirt2 actually has physics that a ATI card can run?

 

[Axil]

Yep - that's the impression I get AMD are aiming more at the support side of things. Having many fingers in the pie and many different pies rather than trying to own one market which is the impression nvidia give.
In the case of nvidia it feels like they are aggressively trying to push this agenda in one area at the expense of their other businesses and if they're not careful (and history had recently proven they don't have the current nuance to do so) they will seriously nark off one of the big players (as if they haven't already) and get seriously burned

 

[NickK]

Someone wanna correct me here plx.

So ATI cards are already physics capable but only in OpenCL?

WHat about old PhysX games? They will have to be ported to OpenCL for ATI cards to ever really work with this?

Dirt2 actually has physics that a ATI card can run?

OpenCL is, at it's heart, just a way of performing A = B*C+D (the operation multiply-and-add, MULA), so is CUDA, OpenGL and DirectX 8,9,10,11,...
All they do is use the MULA to calculate physics rather than shading.
What I'm attempting to say - at the most basic mathematics there's no difference between AMD and nVidia.

Can AMD cards do physics too? Yes. Can they run nVidia's Physx? No because nVidia have made it nVidia compatible only so they can sell their own graphics cards.

nVidia will continue to push their proprietary CUDA. nVidia drivers support OpenCL already (my macbook pro has a nVidia GPU and I have had OpenCL running on it) but only because Apple forced them to. Personally I think CUDA's days are numbered against OpenCL and Microsoft's DirectX.

It's up to nVidia if they code Physx to work with OpenCL or MS. However game development companies won't code for Physx while it reduces their possible sales to nVidia only..

 

[Pendu]

NinkK do you think the solution to cinema and games physics lies with one implementation, considering cinema don't require real time rendering, i thought that is why havok is more or less seen in pc titles and not in cinema effects.

 

[NickK]

NinkK do you think the solution to cinema and games physics lies with one implementation, considering cinema don't require real time rendering, i thought that is why havok is more or less seen in pc titles and not in cinema effects.

The honest answer - I don't know. I'm not an expert in these engines but I'll have a quick look..

If the long term goal is to create a interactive movie, or even a second-life-eqsue MMORPG world as a "movie going experience" rather than just a flat screen then this is a very long way off. Perhaps 30 years before the film companies get to a stage that they understand the best way of making money out of it (it takes a while for artists, directors, actors, producers etc to get happy enough to risk an entire project on it, although Pixar-style movies are leading the way).

Physx and Havok (was well as Bullet) aren't really on the same league as other modelling systems that effects houses such as Industrial Light and Magic or Pixar have created (I see additional kinematics being added features). These tend to be a mix of their own software based on their expertise with a bespoke deployment of hardware in a configuration that suits their hardware.
So I don't see Physx/Havok/Bullet being a threat. It would be interesting to see if Apple get Pixar to port Renderman to OpenCL and then bring some of that to games developers as a Pro SDK... (renderman is OSX 10.5.x not 10.6..)

What is interesting is that Intel have bought Havok, giving it a future but also AMD is now working with Intel to enable hardware acceleration using OpenCL too - gut feeling is that if this works then the next generation Xbox could be an Intel CPU, AMD GPU with Havok running within the MS SDK.. as the Xbox360 has about 5 years left of life, I suspect that the next generation Xbox is well underway in terms of physics support... it's interesting that both Intel and AMD GPUs now support high numbers of SIMD processing (rather than SPMD) but maybe that me jumping to conclusions about Intel's and AMD's next generation of CPUs..
Getting back to the point - don't you think that AMD is partnering OpenCL for physics just as their internal research is probably working hard with Intel/MS and others on the next generation of console... is more than a coincidence?

Ok.. I'm just milling around ideas but I think it has a long way to go..

To me the next jump for us consumers will be when the next generation of consoles appears and the games developers have physics inbuilt deep into the game itself as part of the console OS SDK. Until then I think we'll be stuck with cloth, smoke particle effects and simple stick man rag-dolls.

Question is - is nVidia working with Google for anything? Gut feeling says no due to the culture of the companies.. (ie will the rumoured Google Android/nVidia Tegra partnership result in further projects? Google games console?)

 

[Pendu]

With to regards to the google/nvidia question, perhaps a hand held device could certainly be a possibility. A full fledged console is something i very much doubt.

Also AMD's approach seems a lot more cautious than pioneering. In this instance i prefer the cautious approach, because the pioneering one is bound to lock people out and penalise the consumers who don't have this and that, but the other.