I think he was just making reference that they both have SIMD/vector capability.

Quote:

Original post by Kylotan Quote: Original post by TheFez6255 The main problem with the card is that its not related to one of our senses (not directly anyway). With graphic cards, it improved visuals within the game. With sound cards it was better sound quality (although I know plenty of people that still do not own one). But Physics? There isn't really any big difference to the quality of the games when the average consumer plays your game. As graphics get more photorealistic, players become more attuned to unrealistic behaviour in the in-game models. On this level, better physics will make the gameplay appear more believable. As for sound cards, you're missing the point slightly - the onboard sound today is based on what used to be add-on cards 10 years ago, therefore the add-on card market for it has raised the bar for everybody. I remember the days before digital sound playback was common on PCs. The same thing applies to on-board video, really.

But look at it from the consumers point of view: How will the card make for better games?

Whether or not this eventually gets integrated in to some computer in the future is besides the point. I'm talking strictly from a consumer stand point as to how much the card will improve their gaming experience. My guess is that consumers simply wont get it or care. I'm guessing that most of them are going to say: "I have a computer thats about 3.5 ghz that plays any game I buy perfectly fine. In fact I have physics simulations that run on my machine perfectly fine. Why do I need this card?"

From a developer stand point the card might be gold (although I'm not willing to pay the amount that they want) but to the average gamer, I just don't see them wanting it. I mean I've bought games that are optimized for the card and they ran fine on my machine so why should I get the card (Once again I'm talking from a consumer's point of view)?

To me that means, as a developer, that I have to cut down on a lot of the physics in my game so that it runs on those systems without the card. Personally, unless a game comes along that requires the card, I just don't see it becoming that common place with the average gamer (or at least not in the next 5 to 10 years).

A small question, will the card be practical MMORPGs or any other MMO ?
Consider the procesing power that can be simply replaced by simulating on PPU and even extend it, while CPU deals with player-simulation comunication ... this will alow to simulate large worlds on a single server, while player side simulates software version of engine for small area around him only.

Right now these cards are rather expensive or only come with high end PCs were the 'little' extra money does not weight in much anyway, but what if the price could drop to a $100 something (optimized manufacturing, ..., hardware prices drop) people might think, well I can spend $200 on getting that tuned super high end CPU for 10% better performance or I can get a PPU for half the price that delivers the same overall performace gain (given of course that physics is increasingly used in future games)? A lot of people obviously spend a hell lot of money for a few more frames per second, why not on a PPU if it offers the same.
The fact that the SDK does a good job at abstracting the underlying hardware making it rather easy for developers to support the hardware - as an option - and as the SDK in that case comes virtually for free I think there is a good chance that we see a lot of nice games based on the Ageia technology in the near future. I also think that the additional power of a PPU will mostly be used for special effects physics that has no or little impact on gameplay so that if you should not have a PPU you may still have a good experience. Same thing with the Havok FX approach that is clearly targeted towards special effects. Bottom line, I have some doubts but I see some chances as well, we'll se soon enough.

I couldn't resist throwing in a few pieces of my thoughts as I was reading through the whole discussion.

I guess the problem comes down to specialization vs generalization. There's a reason why there's a specific tool for every task you do everyday. A specialized tool will always work better than a generalized tool when it comes to a task that the specialized tool is designed for. Sure, you can probably use a pocket knife as a screwdriver or a number of other things, but the truth is, the job is better done with a screwdriver.

Since this is in the AI forum, we can look at AI algorithms. When faced with a new problem, where you have no clue what the solution you're looking for may look like, you always start with a very general approach to kind of poke and prod at the search space for a good solution. Then, once you have a good idea of what solution criterions you're looking for, you start targetting those characteristics to do a more specific search and develop more specific solution techniques. These specialized solution methods will always outperform the general methods, unless, of course, there are cases where the problems are so complex that there are no good specialized methods to tackle them.

I guess my point is that specialized hardware will always win in the end. It might be nice if the GPU can do physics, but truth is, its hard enough to make it do anything else, just because its so specialized in doing Graphics. Its not impossible, just that its kind of like fitting a square peg into a circular hole, the fit will never be quite right.

Also, instead of not being sold on the physics processor, I'm not sold on the idea of multi-core being the ultimate solution. Isn't multi-core brute force as well? Because I feel the whole multi-core trend is the off-shoot of the fact that processing speed and performance really isn't increasing as fast as the industry is saying. If you do some research on the Moore's law, the truth is, it hasn't held in a long time. Heck, it never really held true in the first place. Its the industry that's been changing the standards to make it seem true. Its like having an expectation and then tweaking your standards and data interpretation methods to support it. So, multi-core is basically the industry's solution to the problem they face. And what does multicore do for us? Each processor isn't that much faster than before, its just that now, to harness the full power, the developers have to wonder how to actually get the whole multi-threading thing right such that all the cores are utilized properly. Its like the hardware industry kicking the ball into the software developer's court and saying, "we made faster CPUs, but you guys will need to jump through a few hoops to utilize it properly." No, just forking processes (which is what I thought at first, until I really thought about it) isn't the right answer. If multicore really were the answer, we'd see the sound card and GPU dead in a few years as the CPU does everything, but we all know that's not going to happen.

My outlook is that the whole PC industry will split. You'll start having PCs that move closer to being gaming oriented "consoles" and then the ones that are for general purpose. Its pretty much the whole "no free lunch" thing. Heck, its happening already.

But ultimately, will the physics card gain popularity and be picked up by mainstream consumers? Probably won't know till 2 - 3 years down the line, but I'm pretty sure its here to stay, in one form or another. And we will probably see the day that a motherboard will come with integrated GPU, APU (audio), and PPU.

Quote:

Original post by WeirdoFu I guess the problem comes down to specialization vs generalization.

Exactly.

GPUs might be the ultimate example of specialization, but look what happens, they're getting more and more programmable. That's generalization!

The fact of the matter is that too much specialization means limitation. At first OpenGL and Direct3D only offered fixed-function processing and little control was left to the programmer. So every game looked almost the same (looking purely at graphics, not gameplay or art). Nowadays we get programmability at the vertex and pixel level, but you're still limited to rasterized polygonal geometry. Direct3D 10 will allow even more programmability and control, and the evolution won't stop there. Eventually the GPU might become nothing more than lots of SIMD units, controlled by a general purpose CPU, so we can implement even a real-time raytracer or voxel renderer or anything else we can think of.

The PPU will likely have the same evolution. First it will be limited to brute-force processing of objects with simple constraints and forces, behind a closed API. This will look cool first but soon everybody will be bored with the same kind of explosion effect and ragdolls. Developers will want programmability so they can use the PPU's processing power for their own ideas. But this will start looking an aweful lot like a multi-core CPU (in fact it already looks like Cell), or like a highly programmable GPU. In my opinion the kind of processing required for physics doesn't justify the specialization. I'm still convinced the task can be handled perfectly by either the next-generation CPU or GPU. Which of these doesn't matter (since everyone has them), but the PPU won't win.

Quote:

If multicore really were the answer, we'd see the sound card and GPU dead in a few years as the CPU does everything, but we all know that's not going to happen.

Texture sampling and rasterization still requires specialization, so the GPU will most likely stay. But sound processing on the CPU is already done you know, with AC'97 codecs. There's still an audio device but it's almost only to convert the digital signal into an amplified analog signal. Physics processing doesn't require highly specialized hardware like a texture sampler, just generic SIMD units. CPUs already have that and the future generation (Core, Cell) has multiple SIMD units at high clock frequency. Don't forget quad-core and octa-core Intel/AMD processors are already in the making. And with Direct3D 10 and unified architectures we also have lots of programmable SIMD units available on the GPU. I just can't see how PPUs can win this battle. It requires huge investments to use the latest chip technology to make them affordable for a large enough public to catch on.

So to conclude this I think the PPU can have a bit of succes first, just enough to show the importance of good physics in games, but it will quickly get serious competition from software or GPU-based solutions.

No matter what, to make better games with better physics/graphics/whatever you will isolate some audience. Ageia's suggesting an upgrade for people with systems by offering a simple card that can be inserted into a PC. The other options are upgrading you processor, which is usually harder (due to pin configurations and motherboard support) and more expensive, or upgrading your GPU (afterall, to keep "next-gen" graphics and do physics processing that actually affects gameplay, you'll need a really high end card). Fact is some people will not be able to play some games at any time. It's up to developers and publishers to decide. If a bunch of games come out with massive improvements through using hardware, it could quickly become the gamer standard to have the PhysX chip.

The Future of PC Gaming – The Possibilities of Direct3D 10

They're doing physics and even some A.I. on the GPU. A PPU would actually have trouble keeping up with that, and it's clearly a big waste of money. 'Nuff said.

Quote:

Original post by C0D1F1ED The Future of PC Gaming – The Possibilities of Direct3D 10 They're doing physics and even some A.I. on the GPU. A PPU would actually have trouble keeping up with that, and it's clearly a big waste of money. 'Nuff said.

If the D3D 10 API and accompanying hardware is that powerful then I would rather them concentrate on graphics than simulating collision detection or AI. With that type of power you could have actual raytracing (?) and very detailed physics. From that perspective, the PPU is still welcomed [smile]

Quote:

Original post by C0D1F1ED The Future of PC Gaming – The Possibilities of Direct3D 10 They're doing physics and even some A.I. on the GPU. A PPU would actually have trouble keeping up with that, and it's clearly a big waste of money. 'Nuff said.

So let me get this straight your impling that hardware specialized for graphics work, that is already being pushed to it's limits in current games (see oblivion), will run physics better than piece of hardware specialized for physics that isn't doing anything else.

What'll kill the PPU is lack of games and if the first version of it isn't a big enough jump in speed, Since this is the first generation PPU, I expect there to be a lot of optimization room in the drivers and in the chips design. I'll probably wait for 3rd gen before getting one. Unless some killer apps start coming out that take advantage of it before then.