NVIDIA Provides Physics Technology for PLAYSTATION 3

[DarkMatter]

COLLADA is not a physics engine, PhysX is. Even OpenGL or DX11 don't make PhysX obsolete. That's a common missconception. A physics engine is no different than a game/graphics engine. You have D3D, but you still need to make an engine and not everybody knows, wants or can afford to make their own. Also making an engine is relatively easy compared to making a good/efficient engine. Most developers just license it from other companies or more recently some publishers (and not developers as in the past) create their own unique engine to use in all their games. When it comes to physics, almost invariably they license either Havok or PhysX.

 

[LifeOnMars]

yay, more stuttering framerates for the ps3. Lucky people.

 

[Wile E]

yay, more stuttering framerates for the ps3. Lucky people.

Most games don't stutter. Those are only games that are poorly coded, due to lazy developers. Not the fault of the hardware in use.

 

[LifeOnMars]

So its not true that alot of people have problems coding on the machine? Hence bad performance.

 

[Wile E]

So its not true that alot of people have problems coding on the machine? Hence bad performance.

Not really anymore. It was true when it first launched, but even then there were games that ran perfectly. The only ones that run poorly now are the ones from crappy developers.

 

[LifeOnMars]

Thats good then, lets hope the physx is implemented correctly. I have played some games on Playstation 3. I remember I loved Motorstorm.

 

[leonard_222003]

Yes much much better than in a Quad. Remember the GFlops I gave? Well those are not the advertised ones. AFAIK IBM/Sony advertised it as being 256 GFlops. 150 is the maximum sustained Flops found in the chip, I think and 100 was the median low (when SPEs are used). It is commonly accepted that a Core2 has 8-10 Gflops per core depending on the clocks, more when overclocked. We could say that a heavily overclocked Quad has like 50 GFlops, versus the 150 Gflops on the Cell processor. The difference is that Cell's power can only be used in heavily parallel number crunching tasks (such as physics) while Core2 can run everything. Think of it as having 7 workers and an engineer, versus 4 engineers. 4 engineers are much slower doing the hard work, but all of them know what needs to be done at any given time. The 7 workers completely depend on their engineer's instructions.

I can't say you are wrong in what you say but i'm unsure about what IBM says about their CPU , it's an old arhitecture from the days of pentium4 and athlonx2 if i'm not wrong and i know there is theoretical flops number and the one tested in reality and there is a hugeee difference bettwen them.
I was looking trough some forums and i found this
http://img116.imageshack.us/img116/8032/flops1ue4.jpg
a quad QX9770 has 35 real Gflops and a cell processor i have to believe it has 150 gflops ?, i know it's specialized for games and can't do anything but it can do a game menu and AI and has some operating system to run so a game puts that CPU trough a lot of different things to do.
It's just hard to believe that the cell CPU is that good , if it was that good it could've beaten all the CPU's from that time and rule the microprocessor bussines , that stated computing power is pure bullshit for gullible people to believe , Intel and AMD can't exagerate because the next day a happy man makes a software and tests the cpu and shows it can do shit or it shows all that power is not present in some thing he does , IBM had the nerve to state such big numbers because they can't be checked that easily.
People still say even today that PS3 is such a powerfull hardware and can do so much more but no one acces the "power" like it should , or some other excuse , well years have passed and PS3 didn't impressed anyone with great graphics or compiling/encoding/physics or whatever , programers had enough time to learn the "power" of the PS3 and didn't find anything extraordinary in it , even worse for the PS3 the xbox360 wich has worse computing power has better game graphics and the same physics , so where is all that power no one could find IN YEARS ? i think it's in marketing bullshit.

 

[h3llb3nd4]

nVidia+PS3=$$$$+more$$$$
$$$$+more$$$$=no money
no money=no more nice i7

 

[Wile E]

I can't say you are wrong in what you say but i'm unsure about what IBM says about their CPU , it's an old arhitecture from the days of pentium4 and athlonx2 if i'm not wrong and i know there is theoretical flops number and the one tested in reality and there is a hugeee difference bettwen them.
I was looking trough some forums and i found this
http://img116.imageshack.us/img116/8032/flops1ue4.jpg
a quad QX9770 has 35 real Gflops and a cell processor i have to believe it has 150 gflops ?, i know it's specialized for games and can't do anything but it can do a game menu and AI and has some operating system to run so a game puts that CPU trough a lot of different things to do.
It's just hard to believe that the cell CPU is that good , if it was that good it could've beaten all the CPU's from that time and rule the microprocessor bussines , that stated computing power is pure bullshit for gullible people to believe , Intel and AMD can't exagerate because the next day a happy man makes a software and tests the cpu and shows it can do shit or it shows all that power is not present in some thing he does , IBM had the nerve to state such big numbers because they can't be checked that easily.
People still say even today that PS3 is such a powerfull hardware and can do so much more but no one acces the "power" like it should , or some other excuse , well years have passed and PS3 didn't impressed anyone with great graphics or compiling/encoding/physics or whatever , programers had enough time to learn the "power" of the PS3 and didn't find anything extraordinary in it , even worse for the PS3 the xbox360 wich has worse computing power has better game graphics and the same physics , so where is all that power no one could find IN YEARS ? i think it's in marketing bullshit.

He wasn't giving you stated performance numbers. He was giving you real-world tested numbers. You can install Linux on the PS3. It's very easy to get a real-world measurement of the Cell's processing power.

And the Cell cpu can't take over the cpu market because Windows doesn't work on it's architecture.

And the PS3 has impressed many with it's computational power, just not the gamer crowd. It's actually used by quite a few in computing farms to get super-computer performance with a fraction of the cost.

Cell has proven itself to be capable of crunching numbers far better than ANY x86 based cpu. The only reason it is not more popular is because A.) It doesn't run Windows, and B.) It's strength lies in Floating point calculations, and not so much on Integer calculations, which is what most programs are written for.

 

[alexp999]

nVidia+PS3=$$$$+more$$$$
$$$$+more$$$$=no money
no money=no more nice i7

I read today, that game developers are demanding Sony to drop their PS3 prices $50-$100 as they are far too expensive against the Wii and 360. They threatened they would discontinue development for the PS3 if Sony didn't because it was becoming unprofitable with the number of PS3 games sold.

 

[h3llb3nd4]

LOL,.Sony got pwned! but thats kinda sucks!

 

[DarkMatter]

I can't say you are wrong in what you say but i'm unsure about what IBM says about their CPU , it's an old arhitecture from the days of pentium4 and athlonx2 if i'm not wrong and i know there is theoretical flops number and the one tested in reality and there is a hugeee difference bettwen them.
I was looking trough some forums and i found this
http://img116.imageshack.us/img116/8032/flops1ue4.jpg
a quad QX9770 has 35 real Gflops and a cell processor i have to believe it has 150 gflops ?, i know it's specialized for games and can't do anything but it can do a game menu and AI and has some operating system to run so a game puts that CPU trough a lot of different things to do.
It's just hard to believe that the cell CPU is that good , if it was that good it could've beaten all the CPU's from that time and rule the microprocessor bussines , that stated computing power is pure bullshit for gullible people to believe , Intel and AMD can't exagerate because the next day a happy man makes a software and tests the cpu and shows it can do shit or it shows all that power is not present in some thing he does , IBM had the nerve to state such big numbers because they can't be checked that easily.
People still say even today that PS3 is such a powerfull hardware and can do so much more but no one acces the "power" like it should , or some other excuse , well years have passed and PS3 didn't impressed anyone with great graphics or compiling/encoding/physics or whatever , programers had enough time to learn the "power" of the PS3 and didn't find anything extraordinary in it , even worse for the PS3 the xbox360 wich has worse computing power has better game graphics and the same physics , so where is all that power no one could find IN YEARS ? i think it's in marketing bullshit.

In general tasks (even the main thread of a game falls here) a Quad has 4 times the hypothetical power of Cell, if all for cores were actually used, which doesn't happen. Kinda , I don't actually know. It's just a bold guesstimate based on the fact that Cell has only 1 general purpose PPC core, while Quads have 4 (x86). But when it comes to number crunching the Cell is a 8 core (1 PPE + 7 SPEs, SPE = ~dumb processor, Rainman Processor) CPU, with almost twice the output per core compared to Core2, so you can see it as a 16 core CPU in those situations.

About programs using the full potential, you have F@H where the Cell is much faster than any conventional CPU, by almost an order of magnitude I think.

In games you won't see them using the full potential, because in games number crunching is not required, unless you run heavy physics or more complex than adecuate AI algorithms. PhysX can change that.

 

[leonard_222003]

Ok , so we established the cell can do physics faster than any CPU , now let's see what it brings to the table , i guess it has to be better effects in games and physics of course , let's see if they can deliver something better than what we seen until now.

 

[ShadowFold]

I read today, that game developers are demanding Sony to drop their PS3 prices $50-$100 as they are far too expensive against the Wii and 360. They threatened they would discontinue development for the PS3 if Sony didn't because it was becoming unprofitable with the number of PS3 games sold.

Good. My PS3 is a POS. Almost no good games at all. We pretty much never use it anymore..

 

[DarkMatter]

Ok , so we established the cell can do physics faster than any CPU , now let's see what it brings to the table , i guess it has to be better effects in games and physics of course , let's see if they can deliver something better than what we seen until now.

If by that you mean released games, then yes they can do it a lot better. Maybe with the PS3 and the multiplatform they can find an excuse to deliver true physics to games, even when AMD and Intel won't aprove. Only reason they are not including amazing ohysics already is because Ati users wouldn't be able to enjoy them thanks to AMD prefering to be Intel's friend than Nvidia's on this.

With Nvidia + PS3 installed base they don't have an excuse to not implement them IMO.

 

[leonard_222003]

Only reason they are not including amazing ohysics already is because Ati users wouldn't be able to enjoy them thanks to AMD prefering to be Intel's friend than Nvidia's on this.

It's understandable for them ( not us ) , Nvidia didn't adopt dx10.1 or tesselation or any technology at which Ati excelled and of course AMD/Ati will never adopt any technology at which Nvidia can excel or charge money for it , not fair for us from neighter Nvidia who halted the evolution of graphics in games at some point but neighter from AMD/Ati who refuses to adopt technology that could make games better.

 

[FordGT90Concept]

Take Mirror's Edge, for example. the game developers have designed around Physx for the PC, so it requires no extra work to enable it on PS3.

Same goes for future games, if they are working on games that will be Physx enabled on PC, they can now enable Physx on PS3 without paying for coding another physics engine.

They have to work from the ground up to recode everything else as well. Converting physics code isn't that much of a hurdle compared to the rest (liking making use of the CELL engine). Valve learned the hard way on that one when trying to port Half-Life 2 to the PS3.

Simply put, there is little to no value of NVIDIA Physx on PS3 unless it shows significant performance gains over COLLADA. Programmers will stick to what they know and very few PS3 programmers know NVIDIA Physx.

COLLADA is not a physics engine...

Bullet is the physics engine on PlayStation 3, Wii, and Xbox 360.

 

[DarkMatter]

They have to work from the ground up to recode everything else as well. Converting physics code isn't that much of a hurdle compared to the rest (liking making use of the CELL engine). Valve learned the hard way on that one when trying to port Half-Life 2 to the PS3.

Simply put, there is little to no value of NVIDIA Physx on PS3 unless it shows significant performance gains over COLLADA. Programmers will stick to what they know and very few PS3 programmers know NVIDIA Physx.



It runs on a PAL. What PAL Sony has in the PS3 is, as far as I know, unknown. It is most likely Bullet.

Developers don't have to program anything different, PhysX is a two layer API, one is developer/user level, and is exactly the same no matter where you are going to run it. The second layer, kind of kernel or compilator (I don't know exactly what it is) mode is platform specific and is Ageia themselves who make the different compilators. Think of it just as if it was Java, is completely platform independient.

It's been long since PhysX can run on Cell, so it's not a problem anyway. And if you read your link it seems that PhysX can indeed run through COLLADA as well as through PAL anyway. As I stated before one thing is the physics support and another one is the engine.

 

[FordGT90Concept]

Regardless, it is basically another option to use instead of COLLADA/Bullet. It may or may not get used. Because there is no indication PhysX would execute on the GPU, there is likely to be virtually no difference between them for the users.

 

[DarkMatter]

Regardless, it is basically another option to use instead of COLLADA/Bullet. It may or may not get used. Because there is no indication PhysX would execute on the GPU, there is likely to be virtually no difference between them for the users.

It's a better and more complete solution. As I understand it, COLLADA physics is not an engine, it's a bundle of simple routines. PhysX offers everything you need for a lifelike physics reproduction, providing you have the hardware to handle them (it also easily scales to a level a common CPU can handle). There are a lot of benefits to 3rd party solutions versus in-house solutions. They can spend more time and money into making the engine, because they are going to sell it to many people and also is cheaper for the developer using it because they are not the only to pay for the solution. If a complete and efficient engine needs $20 millions to be made, and they sell the thing to 20 developers for $2 each they have enough money to make another better engine and everything evolves. And from the developer point of view they have an efficient solution worth $20, for only $2 million. There's also the fact that a programer group completely dedicated to physics will know how to better make them than other programmers.

 

[Haytch]

Dont you miss the R.I.S.K days ?