How PhysX runs on Nvidia and AMD (ATI) systems

[JoCool]

AMD cards have another advantage, in general you can assume that you get more performance for your money. It has less advanced features. People knew about that trade-off when they bought it. They knew their cards could not process particles, flags, wind, cloth, etc. using PhysX, and that the NVIDIA cards had the potential for that.

Whining now about their own descision and asking for SOE to write their own physics engine (lol) or use a worse one or probably none at all (as in, no banners, no more particles, no cloth wind movements, etc) when a set / toolkit that has been developed for years and is already available is just outright ridiculous.

You know what. I'd want my Prowler to sport our future Outfit's flag. I want cloth banners hanging down the bases' walls, trees moving in the wind, water trails, water flashes, leaves bristling, a more alive environment.

Who would not want that? How can you still defend your point of not implementing such features but use mediocre ones?

Developing an own Physics engine could probably take years to even reach the level of the PhysX developer's toolkit.


To anyone who wants to respond to this, just answer yourself the following question: It was your choice, wasn't it?

[Masterr]

Originally Posted by JoCool

AMD cards have another advantage, in general you can assume that you get more performance for your money. It has less advanced features. People knew about that trade-off when they bought it. They knew their cards could not process particles, flags, wind, cloth, etc. using PhysX, and that the NVIDIA cards had the potential for that.

Whining now about their own descision and asking for SOE to write their own physics engine (lol) or use a worse one or probably none at all (as in, no banners, no more particles, no cloth wind movements, etc) when a set / toolkit that has been developed for years and is already available is just outright ridiculous.

You know what. I'd want my Prowler to sport our future Outfit's flag. I want cloth banners hanging down the bases' walls, trees moving in the wind, water trails, water flashes, leaves bristling, a more alive environment.

Who would not want that? How can you still defend your point of not implementing such features but use mediocre ones?

Developing an own Physics engine could probably take years to even reach the level of the PhysX developer's toolkit.

Who said anything about them writing their own physics engine?

[zomg]

I'm not sure how people get the idea that PhysX on CPU is poorly implemented.

Did it ever occur to you, that the tech put into Nvidia GPUs comes from Ageia's card, which was specifically developed to run physics code fast. A processor specifically developed for a single task is much more efficient at it than a general purpose processor like a CPU. This is also why we have distinct GPUs.

So could it be, that because of the advanced physics enabled by the parallel processing power of the GPU and the Ageia tech, it actually allows it to run comparatively better physics models?

And now with that, since CPUs aren't as efficient with parallel processing as GPUs, we can directly draw a parallel to why it would perform poorly on a CPU. Not because of bad drivers, but because of simply the fact that the CPU isn't good enough for it.

So basically, get a better CPU or an Nvidia card.

Feel free to provide reliable facts about PhysX being poorly implemented, and that being the cause, rather than what I said here. I doubt you can.

[EVILoHOMER]

Contrary to some headlines, the Nvidia PhysX SDK actually offers multi-core support for CPUs. When used correctly, it even comes dangerously close to the performance of a single-card, GPU-based solution. Despite this, however, there's still a catch. PhysX automatically handles thread distribution, moving the load away from the CPU and onto the GPU when a compatible graphics card is active. Game developers need to shift some of the load back to the CPU.


The effort and expenditure required to implement coding changes obviously works as a deterrent. We still think that developers should be honest and openly admit this, though. Studying certain games (with a certain logo in the credits) begs the question of whether this additional expense was spared for commercial or marketing reasons. On one hand, Nvidia has a duty to developers, helping them integrate compelling effects that gamers will be able to enjoy that might not have made it into the game otherwise. On the other hand, Nvidia wants to prevent (and with good reason) prejudices from getting out of hand. According to Nvidia, SDK 3.0 already offers these capabilities, so we look forward to seeing developers implement them.

[julfo]

Originally Posted by JawsOfLife

With no sources cited I have to say you have no credibility with this claim in my book. Nothing personal, it's just I could say "Well actually nVidia has completely optimized PhysX to run well on CPU and their GPU's really are that much better!" and we'd cancel each other out until some actual data came into play.

Cheers

http://www.tomshardware.com/news/phy...ics,10826.html

Interesting read.

[JawsOfLife]

Originally Posted by julfo

http://www.tomshardware.com/news/phy...ics,10826.html

Interesting read.

Thanks for the link bro, I have no problem being proven wrong if it means the truth is promoted

Love Toms, and it is an interesting read. This is a very tricky situation, though. Is purposefully nerfing the CPU potential of PhysX so that an nVidia GPU is the best option by far to play the games on very ethical conduct? Decidedly not. Would nVidia cards be so compelling if the CPU performance more closely matched it? No it wouldn't. So it is a balance between monopolizing great GPU physics performance and selling more cards while still not screwing over everyone who has other brands. It is a tough call, and I don't see a clear solution to the problem. Why would nVidia upgrade the code to x86 if it would lose them money? They wouldn't, that's a silly concept. And again, PhysX IS open source for developers to optimize, if they so desire.

[julfo]

Originally Posted by JawsOfLife

Thanks for the link bro, I have no problem being proven wrong if it means the truth is promoted

Love Toms, and it is an interesting read. This is a very tricky situation, though. Is purposefully nerfing the CPU potential of PhysX so that an nVidia GPU is the best option by far to play the games on very ethical conduct? Decidedly not. Would nVidia cards be so compelling if the CPU performance more closely matched it? No it wouldn't. So it is a balance between monopolizing great GPU physics performance and selling more cards while still not screwing over everyone who has other brands. It is a tough call, and I don't see a clear solution to the problem. Why would nVidia upgrade the code to x86 if it would lose them money? They wouldn't, that's a silly concept. And again, PhysX IS open source for developers to optimize, if they so desire.

No problem And thank you for constructing an intelligent response. I know many people (including myself) would have flipped out. It's refreshing to find someone who doesn't.

I think you've more or less hit the nail on the head here. nVidia are walking a fine line right now between making their product more desirable (physX wooo!), and making it something that the general populace despises (because of what could be viewed as underhand tactics). At the moment they're doing an admirable job of it.

Right now I'm using AMD rather than nVidia. I bought a 7970. The thing that made me pause before making this decision was physX. nVidia are clearly succeeding to some degree.

It is a difficult situation, and there aren't any clear solutions. As you said, nVidia are unlikely to upgrade to x86, it would be silly for them to do so. At the end of the day, it has to be said that really the onus is on the developers of the game to use physX to the best degree. It is possible, with some work, to convert physX to x86/SSE, and assuming the developers implement threading properly, there should be negligible difference between GPU and CPU (as the post above nicely demonstrated).

Even if they don't (although I have the utmost confidence in the Dev Team) there is still another solution for most AMD users: Hacked nVidia drivers with a dedicated nVidia card for physX offloading. Using the modified 1.05ff drivers one can create a "hybrid" set-up which is comparable to a single card nVidia set-up. I personally don't think physX is anything to worry about. Others disagree, and they're welcome to.

[JawsOfLife]

Originally Posted by julfo

No problem And thank you for constructing an intelligent response. I know many people (including myself) would have flipped out. It's refreshing to find someone who doesn't.

I think you've more or less hit the nail on the head here. nVidia are walking a fine line right now between making their product more desirable (physX wooo!), and making it something that the general populace despises (because of what could be viewed as underhand tactics). At the moment they're doing an admirable job of it.

Right now I'm using AMD rather than nVidia. I bought a 7970. The thing that made me pause before making this decision was physX. nVidia are clearly succeeding to some degree.

It is a difficult situation, and there aren't any clear solutions. As you said, nVidia are unlikely to upgrade to x86, it would be silly for them to do so. At the end of the day, it has to be said that really the onus is on the developers of the game to use physX to the best degree. It is possible, with some work, to convert physX to x86/SSE, and assuming the developers implement threading properly, there should be negligible difference between GPU and CPU (as the post above nicely demonstrated).

Even if they don't (although I have the utmost confidence in the Dev Team) there is still another solution for most AMD users: Hacked nVidia drivers with a dedicated nVidia card for PhysX offloading. Using the modified 1.05ff drivers one can create a "hybrid" set-up which is comparable to a single card nVidia set-up. I personally don't think PhysX is anything to worry about. Others disagree, and they're welcome to.

I totally agree with your choice. I have a GTX 460 SE, purchased last year, because I am a student and had a strict budget to build a computer with. However, I feel for the $130 I paid for it, it was a great deal. But I have really been diving in to benchmarks recently and seen the deficiencies of nVidia's GPUs. Compared to AMD, they are a bit of one-trick ponies. Except the one trick they do, gaming, is done incredibly well. Their cards do tend to choke at high resolutions with high AA, but on the whole do very very well at the 1080p point and higher with lowered AA, which is where most of the gaming community is at right now. AMD has lower max fps in general but performs much better at more extreme settings. On top of that, their GPUs also have much better compute performance in the majority of cases (SHA 256 hashing and many software compute functions come to mind). nVidia pretty much only wins in CUDA, and PhysX.

As a last note, while it is possible to mod in PhysX support, I do not see the majority of AMD users doing that. It is so niche (and possibly warranty-voiding) that most won't attempt to do it, though that doesn't make it an invalid tactic.

[zomg]

Originally Posted by julfo

http://www.tomshardware.com/news/phy...ics,10826.html

Interesting read.

Interesting indeed, but I wonder how dated that is. That article is from 2010, and it talks about PhysX 3.0, which was released last year, so it could be it has changed towards better since then.