The physics "claims" thread...

[Aristotelian]

http://www.youtube.com/watch?v=-x9B_4qBAkk

Lonbjerg: I'm not sure how your response to my last post addressed BFG's point (that non PhysX games can have great physics in them also, so PhysX isn't the second coming of physics in computer games0.

Regarding the link I'm posting here, I tend to think some of the effects are overdone - exaggerated just to make them a selling point. I don't see PhysX as a huge deal (yet).

 

[Lonbjerg]

You are the one making false claims. We all know and understand that scripted destructible environment means "debris dissapears into the ground, into thin air....or remains static." GPU physics serves to improve upon that... but the improvement is NOT Significant enough to matter so they cheat by crippling non GPU physics implementations on select games.

So you are comparing apples to oranges, but demand they tax the system the same?
And I had no problems playing MAFIA2 with GPU physics...so once again...what is your point?

 

[TransistorsX]

This thread clearly shows that the OP is nVidia biased and instead of posting facts, it is only posting marketing spins and agressive attacks against posters who can't agree with him. I own an nVidia and AMD card and I really experienced PhysX on all GPU accelerated games, and yet you haven't even played Monopoly or posted something meaningful and usefull. I never said that running PhysX on CPU resulted on playable frame, read carefully my previous post before jumping on conclusions, as you will be the one on bad light.

 

[Skurge]

I don't think PhysX is crap and appreciate what nVidia is trying to do. I think the PhysX SDK is very popular with developers and their tools are impressive and one can make the argument that nVidia is the leader in Physics; as some can make the argument that Havok is the leader.

To me, this is all good, because Physic talk is still a lightning rod discussion -- important, powerful companies have their own strategies to try to bring more Physic content to gamers. For me, it is the chaos that may forge some common ground.

What has been underwhelming to some degree is the content of GPU Physics. Was hoping for one AAA title a month, or 6-12 titles a year. Sadly, this isn't the case and the question is; Why?

Is it the division? Developers don't desire to add GPU Physics?

HavokFX ripped away when it was showing such promise.

Bullet, which sounds great with no content at all.

Just feel so strongly that Physics can do so much to enhance gaming and there just seems to be so many obstacles, hehe!

Am i the only person who thinks we don't need GPU's to run physics? CPU's are being taxed less and less, relative to the increase in CPU power. Look at SW Force unleashed, not a great game, but the physics there are some of the best I've seen. It might not have as many particles as Mafia 2, but it looks way more realistic and it runs on any modern CPU and it isn't even scripted. Its a mix of havok and euphoria. Same goes for Ghostbusters and Far Cry 2. GPUs might be an order of magnitude faster than CPUs, but they already have their hands full.

With clever programing. You can get excellent physics even without the power of a GPU.

 

[taltamir]

So you are comparing apples to oranges, but demand they tax the system the same?

This is just random words that have absolutely nothing to do with anything I said.
Apples to oranges comparison? how the heck does this even come close to relate to what I said?

And I had no problems playing MAFIA2 with GPU physics...so once again...what is your point?

... I clearly and carefully elucidated my point, you keep ignoring it or inventing an easily disprovable misrepresentation of it. Go reread my post.

 

[SirPauly]

Am i the only person who thinks we don't need GPU's to run physics? CPU's are being taxed less and less, relative to the increase in CPU power. Look at SW Force unleashed, not a great game, but the physics there are some of the best I've seen. It might not have as many particles as Mafia 2, but it looks way more realistic and it runs on any modern CPU and it isn't even scripted. Its a mix of havok and euphoria. Same goes for Ghostbusters and Far Cry 2. GPUs might be an order of magnitude faster than CPUs, but they already have their hands full.

With clever programing. You can get excellent physics even without the power of a GPU.

Probably not! To have game-changing and dramatic jumps in realism and fidelity one may need an order of magnitude more performance. The raw potential is there now to do that. Is just doesn't make sense not to try to take advantage of this and why nVidia, ATI originally, they never did STFU back then about it, even Havok, with HavokFX. It has been so political and the powerful players have their own agendas and strategies and the consumer has suffered to some degree in my mind.

 

[Skurge]

Probably not! To have game-changing and dramatic jumps in realism and fidelity one may need an order of magnitude more performance. The raw potential is there now to do that. Is just doesn't make sense not to try to take advantage of this and why nVidia, ATI originally, they never did STFU back then about it, even Havok, with HavokFX. It has been so political and the powerful players have their own agendas and strategies and the consumer has suffered to some degree in my mind.

I don't agree that we need GPUs to have proper game changing physics. Maybe more convincing looking, but not game changing. I bet you EPIC will be able to do that with the updated UE3 engine and will do well on CPUs.

 

[SirPauly]

The only kind of game changing I see with my broken crystal ball is by the CPU 'till there are some sort of standards and the players can somehow play nice and agree on something.

Why would a developer add an absolute, integral game-play physic improvement that can only be enhanced by one IHV? Doesn't make logical sense. However, if there is more openness with robust tools, could easily see developers slowly adding more physic game-play immersion and let their imaginations be the deciding factor.

 

[Lonbjerg]

AMD dosn't think the CPU is all that for Physics:

http://www.bit-tech.net/hardware/graphics/2011/02/17/amd-manju-hegde-gaming-physics/1

.oO( Waiting for a reversal of stance...now that the "brand" has spoken"...)

It seems it' only fans of AMD thinking the CPU is good enough...not AMD...oh the irony :biggrin:

 

[Gloomy]

AMD dosn't think the CPU is all that for Physics:

http://www.bit-tech.net/hardware/graphics/2011/02/17/amd-manju-hegde-gaming-physics/1

.oO( Waiting for a reversal of stance...now that the "brand" has spoken"...)

It seems it' only fans of AMD thinking the CPU is good enough...not AMD...oh the irony :biggrin:

Are you retarded?


There is nothing positive or constructive to come from openly inquiring into the mental faculty of the poster you are addressing.

This is a technical forum, you are expected to refrain from posting personal commentary that is negative, insulting, or generally disrespectful.

Debate the merits of the contents of the post, not the irrelevant background material relating to that of the poster.

(for deftly handled examples of how to do this you need look no further than to the following 3 posts contributed by your fellow forum colleagues)

Moderator Idontcare

 

[Arkadrel]

AMD dosn't think the CPU is all that for Physics:

http://www.bit-tech.net/hardware/graphics/2011/02/17/amd-manju-hegde-gaming-physics/1

.oO( Waiting for a reversal of stance...now that the "brand" has spoken"...)

It seems it' only fans of AMD thinking the CPU is good enough...not AMD...oh the irony :biggrin:

..Manju Hegde, a new start-up called Ageia ambitiously launched its dedicated PhysX accelerator.

last year Hegde was poached by AMD to join the Fusion team, prompting all sorts of speculation about AMD's plans for competing with PhysX.

So one of the guys that made Ageia, with its PhysX, now works for AMD working on makeing a PhysX version for its APUs?

That doesnt sound like its a bad thing, does it? I mean maybe we ll see the IGP on the cpu work like a discrete physx card, when you plug in another GPU?

Bullet Physics is capable of creating impressive rigid body physics effects, and has already been used in big-name blockbuster movies such as 2012, Hancock and The A-Team. It's also already made it into several games, including Grand Theft Auto IV, Red Dead Redemption and Toy Story 3.

Bullet Physics, sounds like it ll have a better chance of getting adopted, simply because any GPU will be able to run it.

 

[Outrage]

AMD dosn't think the CPU is all that for Physics:

http://www.bit-tech.net/hardware/graphics/2011/02/17/amd-manju-hegde-gaming-physics/1

.oO( Waiting for a reversal of stance...now that the "brand" has spoken"...)

It seems it' only fans of AMD thinking the CPU is good enough...not AMD...oh the irony :biggrin:

AMD's physic team is involved with Fusion. AMD's vision is to run the physics on the Fusion apu and let the graphics card do the graphics.

When this happens, amd will have an advantage over Intel.

 

[Lonyo]

AMD dosn't think the CPU is all that for Physics:

http://www.bit-tech.net/hardware/graphics/2011/02/17/amd-manju-hegde-gaming-physics/1

.oO( Waiting for a reversal of stance...now that the "brand" has spoken"...)

It seems it' only fans of AMD thinking the CPU is good enough...not AMD...oh the irony :biggrin:

Glorious demos have been shown off, cash has been splurged and big promises have been made, but the truth is that really only a handful of PC games actually feature hardware-accelerated physics. What's more, those that do, such as Mafia 2, mainly use it for particle eye-candy, rather than changing the actual game mechanics.

http://forums.anandtech.com/showpost.php?p=31343460&postcount=156
CPU physics.

1) GPU physics hasn't been used for much.
2) CPU physics can do more for the underlying game than it has currently been used for, including on consoles.

So, why focus on GPU physics when we haven't even fully exploited the full potential of more widely support CPU physics, when there isn't any kind of universal accelerated (non-CPU) method of adding physics to games at the moment?
It's more sensible to either 1) use CPU physics in a game-changing way or 2) work with an open and widely supported way of bringing better/faster physics to gaming than caring about a limited use technology (PhysX on an NV GPU).

AMD is going for #2.
NV is going for #crap.

 

[Skurge]

AMD is going for #2.
NV is going for #crap.

LOL. This is funny in so many different ways.

 

[waffleironhead]

The whole scripted versus "real" physics arguments in this thread remind me a lot of the phenom release era, when I heard a lot of "but, its not a REAL quad core" from the amd camp trying to downplay intels offerings.

I'll take scripted and real enough, versus the way the "real" stuff has been implemented so far.

 

[SirPauly]

http://forums.anandtech.com/showpost.php?p=31343460&postcount=156
CPU physics.

1) GPU physics hasn't been used for much.
2) CPU physics can do more for the underlying game than it has currently been used for, including on consoles.

So, why focus on GPU physics when we haven't even fully exploited the full potential of more widely support CPU physics, when there isn't any kind of universal accelerated (non-CPU) method of adding physics to games at the moment?
It's more sensible to either 1) use CPU physics in a game-changing way or 2) work with an open and widely supported way of bringing better/faster physics to gaming than caring about a limited use technology (PhysX on an NV GPU).

AMD is going for #2.
NV is going for #crap.

Because of the raw potential and performance benefits over the CPU. Why have GPU Processing at all and simply let's just use CPU's, too. nVidia's method may not be ideal but at least they're trying, risking and spending resources on something they believe in. I don't have a problem because their spending where their mouth is. I don't desire PhysX to die, fail but evolve and mature and to see PhysX be ported to OpenCL some day.

It's easy to sit and judge without accountability and risk -- point fingers but they're spending millions of dollars to try to bring GPU Physics to the PC. I can at least try to learn, to try to understand their point-of-view and try to be fair.

 

[TransistorsX]

http://forums.anandtech.com/showpost.php?p=31343460&postcount=156
CPU physics.

1) GPU physics hasn't been used for much.
2) CPU physics can do more for the underlying game than it has currently been used for, including on consoles.

So, why focus on GPU physics when we haven't even fully exploited the full potential of more widely support CPU physics, when there isn't any kind of universal accelerated (non-CPU) method of adding physics to games at the moment?
It's more sensible to either 1) use CPU physics in a game-changing way or 2) work with an open and widely supported way of bringing better/faster physics to gaming than caring about a limited use technology (PhysX on an NV GPU).

AMD is going for #2.
NV is going for #crap.

Lol I had a small burst of laugh waiting in the bank when I saw that LOL



evolucion8

Go away. You are not welcome here.


esquared
Anandtech Forum Director

 

[SirPauly]

So one of the guys that made Ageia, with its PhysX, now works for AMD working on makeing a PhysX version for its APUs?

That doesnt sound like its a bad thing, does it? I mean maybe we ll see the IGP on the cpu work like a discrete physx card, when you plug in another GPU?




Bullet Physics, sounds like it ll have a better chance of getting adopted, simply because any GPU will be able to run it.

I'm all for it --- any advancement toward improving Physics is neat.

 

[Skurge]

Because of the raw potential and performance benefits over the CPU. Why have GPU Processing at all and simply let's just use CPU's, too. nVidia's method may not be ideal but at least they're trying, risking and spending resources on something they believe in. I don't have a problem because their spending where their mouth is. I don't desire PhysX to die, fail but evolve and mature and to see PhysX be ported to OpenCL some day.

It's easy to sit and judge without accountability and risk -- point fingers but they're spending millions of dollars to try to bring GPU Physics to the PC. I can at least try to learn, to try to understand their point-of-view and try to be fair.

The GPU is already busy doing the graphics. Until we have a surplus of GPU performance, I still think its better done on the CPU which we do have a surplus of power.

 

[Lonyo]

The GPU is already busy doing the graphics. Until we have a surplus of GPU performance, I still think its better done on the CPU which we do have a surplus of power.

Man, you sound exactly like some random company I know of.

I think they are called... what's the name.

Need a minute to remember.

I'll get there in the end.

NVIDIA! I remember now!

They called it... the Optimized PC.

Apparently it's better to buy a faster GPU because getting a faster CPU doesn't improve your gaming performance, because games are typically GPU bound.

http://news.cnet.com/8301-17938_105-9916075-1.html
http://www.nvidia.com/content/opc/v2/widget_code.asp?lang=en-us ->
http://www.nvidia.com/content/opc/v2/OptimalPlatformEnthusiast.swf

Don't overbuy the CPU, because that's not where the performance limit is, get a better GPU because that will improve your gaming more, because games are GPU limited. - Nvidia, 2008.

Buy a second GPU so that you don't need to use your primary GPU (which is already the limiting factor) to do physics calculations. The CPU can't do anything. - Nvidia, 2008-2011.

Physics is great for all platforms and we are working hard on it, but it's not multi-core aware by default so it's seriously limited on multi-core systems. - Nvidia, 2010.

Anyone would think they are deliberately trying to make people underbuy CPUs, inhibit CPU physics performance, and persuade everyone to increase the load on their existing GPU and buy more GPUs and faster GPUs.
It's called business. And it's not typically intended to help the consumer on the whole.

 

[3DVagabond]

You have a problem though.
On the 4c/8T CPU, you spend almost as much time waiting for physics threads to finish as you do comupting them.

It's abot being parrellel and 8 thears is no match for 512 CUDA cores...do the the math.

Or read claim #1...with documentation saying otherwise from Bullet Physics.

It's really sad that myths are being used as facts.

How many of those 512 CUDA cores have nothing else to do? Most games are only using 1 or 2 cores of your CPU. The rest are free to do physics. CPU's are rapidly increasing in parallelism. Games, where the CPU are concerned, are not. At the rate PhysX is going, we'll have plenty of excess CPU cores for physics before we see any major inroads with PhysX. How many years has it been around now, and look at how imbedded it is. Not very.

 

[wuliheron]

It's a moot argument either way.

Consoles and graphics drive the development of games and the current generation of consoles suck for physics. The next generation should be much more capable with fast quad core processors and modern graphics cards. That will put pc gamers right back where we've always been requiring both a strong processor and graphics card to run games that are still developed to take advantage of both.

Physics and AI run best on 80 or more processors (like in a gpu), but 8 cores is a good minimum due to the symmetry of the equations and multiples of 8 are ideal. It allows for simplified physics that don't need to be run on the gpu but can still produce some wow factor. Like ragdoll physics and other standards in games we should should see an explosion of rudimentary physics standards developed that run well on 8 cores and maybe a little slow on fast quads. Thankfully consoles can get away with fewer processors since the games are optimized for them and they don't run operating systems. The next generation consoles should have fast quad cores with modern graphics cards making them much more physics friendly. Then we'll see some real physics action in games as it becomes significantly easier for developers to implement and moves further up their list of priorities.

In the long run I'd expect 8 core cpu/gpu combinations that can run portable devices. Some you might be able to add a discrete gpu onto for more power allowing the cpu/gpu to do all the physics.