AMD sponsors release of free Open Source DMM Physics engine.

[GodisanAtheist]

http://www.techpowerup.com/117030/A...system_with_Free_DMM_for_Game_Production.html

-Surprised this wasn't posted here already. While this is certainly (yet another...) step in the right direction, there is little reason to celebrate until we see it permeate across the market.

For reference, The Force Unleashed used DMM.

 

[zerocool84]

An open source Physics engine that anyone can use no matter what card you have??? Who would have thought that would be good for the industry?

 

[Patrick Wolf]

I hope this leads to the inevitable death of PhysX more quickly.

 

[tviceman]

An open source Physics engine that anyone can use no matter what card you have??? Who would have thought that would be good for the industry?

I'm all for improving PC graphics, gaming, and the way we play games, but....

"Great, this will just turn off stupid things that you can do with any card/cpu to make you think you need a really powerful GPU. I mean these are things that can be done without gpu physics and will be taken out to sell cards and say that you can't get these cool effects without it even though we know we can cus games have been doing them for years.Also you're right I haven't played any of these games yet. I don't know anyone who has or will cus they'll all be terrible games."

I love how I can change 7 words in three quotes without changing the substance surrounding half your previous arguments and it can be easily applied against your statements now.

 

[zerocool84]

I'm all for improving PC graphics, gaming, and the way we play games, but....



I love how I can change 7 words in three quotes without changing the substance surrounding half your previous arguments and it can be easily applied against your statements now.

And of course you magically leave out half of the quotes that I've said. I've said that if they provide meaningful games and change the way it's play like they said they were going to I'm all for it but of course for some reason you forgot about that but it's ok cus it doens't help your argument at all. See you have to change it to have it mean anything to you since you have no argument. Can't even give you a good try. It's not my fault your Nvidia lords haven't had Physx do anything meaningful at all and given gamers a reason to think it's important.

 

[Lonyo]

I hope this leads to the inevitable death of PhysX more quickly.

PhysX doesn't need to die, it just needs to follow the lead of DMM.

 

[Patrick Wolf]

PhysX doesn't need to die, it just needs to follow the lead of DMM.

That would mean ATI hardware can run it just the same as nvidia hardware. I doubt nvidia would do that since, well, they're a-holes.

 

[BD231]

That would mean ATI hardware can run it just the same as nvidia hardware. I doubt nvidia would do that since, well, they're a-holes.

No they're just arrogant and overpaid, there's a difference ... oh wait that's right no there isn't.

 

[BenSkywalker]

Hopefully they get fluid and cloth support out soon, then things could get quite interesting.

 

[Patrick Wolf]

"On AMD platforms, ATI Stream technology is used to drive the enhanced game experience."

I may interpret this as "it will run like crap on nvidia hardware." If not, what's this mean? I think ATI Stream has something to do with how the CPU and GPU work together so maybe the physics would work "better" on a platform with an AMD CPU and ATI GPU?

 

[NoQuarter]

"On AMD platforms, ATI Stream technology is used to drive the enhanced game experience."

I may interpret this as "it will run like crap on nvidia hardware." If not, what's this mean? I think ATI Stream has something to do how the CPU and GPU work together so maybe the physics would work "better" on a platform with an AMD CPU and ATI GPU?

Stream is ATI's version of CUDA, Nvidia wouldn't be able to run it but they would be able to run the OpenCL or DirectCompute versions, no idea what performance difference would be like.

 

[grimpr]

It will run just fine on new NV hardware, AMD's OpenCL implementation has the additional benefit that code also runs on all available cpu cores if desired.

http://bulletphysics.org/wordpress/

 

[cbn]

I noticed "Bullet physics" specifically mentioned.

So now we are talking wind and bullet drift being factored into the strategy of First person shooters?

 

[Attic]

... AMD's OpenCL implementation has the additional benefit that code also runs on all available cpu cores if desired.

http://bulletphysics.org/wordpress/

You mean they won't artificial limit it to run on only one core to scew performance like physx? This sounds almost like it's good for the gamer, i'm guessing nVidia will try to sabatoge it.

 

[NoQuarter]

I noticed "Bullet physics" specifically mentioned.

So now we are talking wind and bullet drift being factored into the strategy of First person shooters?

Bullet is a product name, like Havok, PhysX, etc.

 

[cbn]

Bullet is a product name, like Havok, PhysX, etc.

That makes sense. Wouldn't bullet drift from wind be cpu based anyway?

 

[blanketyblank]

"On AMD platforms, ATI Stream technology is used to drive the enhanced game experience."

I may interpret this as "it will run like crap on nvidia hardware." If not, what's this mean? I think ATI Stream has something to do with how the CPU and GPU work together so maybe the physics would work "better" on a platform with an AMD CPU and ATI GPU?

I see that as a concealed threat to nVidia. If you want this to work for your card you better start help with the development of this engine or else it will be optimized for ATI tech although it will support nVidia cards.

 

[cbn]

I see that as a concealed threat to nVidia. If you want this to work for your card you better start help with the development of this engine or else it will be optimized for ATI tech although it will support nVidia cards.

Now that this form of GPU Physics is an open standard what are the chances we could see these effects actually implemented in game strategy.

How well could CUDA run this new standard compared to Stream technology?

 

[TheMeanestGuest]

While it's always nice to see things be open source, I'll be more impressed when I see it in game and working, as well as effecting gameplay in a meaningful way.

Of course, just by the inclusion of DMM I imagine that would be more likely, but we'll just have to wait and see.

I'd imagine it will probably be at least a year before we see anything.

 

[BenSkywalker]

You mean they won't artificial limit it to run on only one core to scew performance like physx?

http://www.geeks3d.com/20100304/preview-physx-simulation-on-a-16-core-cpu/

It runs on at least 16 cores. There may be issues with implementation, but that isn't a limitation of the PhysX SDK by any means.

How well could CUDA run this new standard compared to Stream technology?

How well can nV handle general GPGPU code compared to ATi. nV has been working along with AMD to get Bullet support on GPUs for a while now. While ATi worked out a deal to bring DMM to PCs through their sponsorship, Bullet has been coming for quite a while now and has broad support.

While it's always nice to see things be open source, I'll be more impressed when I see it in game and working, as well as effecting gameplay in a meaningful way.

Why does physics require an impact on gameplay to be worth doing? If you want to see some excellent physics modelling in games today, check out some of the top tier racing titles. Extremely accurate physics IMO are going to be very nice for things like foliage that moves instead of clipping through it etc. Suspension of disbelief is the big thing I'd like to see physics help out with. In terms of game changing dynamics, most people point to destructable environment. I had that back in the mid 80s with Super Mario Bros, I'm looking for something a bit beyond that

 

[TheMeanestGuest]

Why does physics require an impact on gameplay to be worth doing? If you want to see some excellent physics modelling in games today, check out some of the top tier racing titles. Extremely accurate physics IMO are going to be very nice for things like foliage that moves instead of clipping through it etc. Suspension of disbelief is the big thing I'd like to see physics help out with. In terms of game changing dynamics, most people point to destructable environment. I had that back in the mid 80s with Super Mario Bros, I'm looking for something a bit beyond that

While I do appreciate enhanced visuals, they are certainly not the most important factor for me in regards to the enjoyment I extract out of my games. Fluttering cloth and interactive dust is nice and all, it's just not where I'd like to see advanced physics implemented when there are other much cooler things that could be done.

I like the physics interactions in many racing games, (I think Grid did a fairly nice job) this contributes to it remaining one of my preferred genres of game, although, obviously a good physics implementation won't make up for a poor game, but this is true of any genre.

I think why a lot of people point to realistic (or at least more realistic) destructible environments is because, well, that is genuinely what they want to see. Like you said, suspension of disbelief is important, and I think that destructibility is one of the best ways to accomplish that, and it also changes the gameplay in a pretty big, and, to me, fun way. I would love to see realistic interactions with foliage in games (as well as landforms and all sorts of other things), but I'd rather see realistic interactions with buildings first. I don't know about you, but I was massively entertained by the destructibility in Red Faction: Guerrilla, I'll admit it could have been better, particularly in density, but it was still quite good.

GPU acceleration of physics is something that I support, and an open implementation I support even more, but I think that all the fluffy aesthetic uses for hardware accelerated physics should probably be done after we get some more substantive stuff out of it.

 

[cbn]

Deleted

 

[GaiaHunter]

Why does physics require an impact on gameplay to be worth doing? If you want to see some excellent physics modelling in games today, check out some of the top tier racing titles. Extremely accurate physics IMO are going to be very nice for things like foliage that moves instead of clipping through it etc. Suspension of disbelief is the big thing I'd like to see physics help out with. In terms of game changing dynamics, most people point to destructable environment. I had that back in the mid 80s with Super Mario Bros, I'm looking for something a bit beyond that

I guess it depends - I guess some people want reality simulators, but having something that is completely impossible in a game isn't exactly bad either. Depends on the game type, on the implementation, etc.

But so far the stuff that is out there isn't exactly astonishing - sure the technical and accuracy aspect is quite interesting, but many things we have seen so far with GPU accelerated physics can be accomplished in different ways.



Physics will most likely take off when it can use all the resources available on someone system, make you open your mouth and have no cost in frame rate.

It is a question of timing and so far, and for different reasons, GPU physics hasn't take off.

Sure, things need to start at some point, and it is starting, but it is far from mainstream atm and we can be excited by the future of it but we (or some of us) aren't exactly excited with the present of it.

 

[Kuzi]

PhysX doesn't need to die, it just needs to follow the lead of DMM.

You mean for NV to allow PhysX to run on OpenCL/DirectCompute? I don't see that happening, but it would be the right move, this way all supporting hardware, ATI's, Intel's etc would also run on PhysX.

If that happens, we'll have more competition between all the physics engines/middleware such as Bullet, Havoc, and PhysX, and developers would have more choice which is great. And I believe certain software developers may build their own physics engines in-house and then license them to others, iD Software comes to mind. I would say Epic too, but they use PhysX now, which is kind of surprising.

 

[Kuzi]

Now that this form of GPU Physics is an open standard what are the chances we could see these effects actually implemented in game strategy.

How well could CUDA run this new standard compared to Stream technology?

It is not so much about CUDA/Stream as it is about the underlying hardware and drivers. In the future there will be cases where we can directly compare NV and ATI cards for GPGPU and Physics capability. Which would be faster? If we take Fermi for example, a GPU which was designed from the ground up for these tasks in mind, it won't be too surprising if it is x2 faster than the 5000 series cards at certain GPGPU tasks.