Dedicated Physics Processors: Yay or Nay?

Some rather amusing news bouncing about with those PPUs.

John Carmack has said what had to be said: That they are useless, and that having multiple processing cores manages the same thing fine. (With that growing to "good" as processor technology improves in every field instead of just one pointless field that is game physics).

Ageia shot back with an absurd claim to make their processors look worthwhile, that there are timing issues between multiple cores when they are being split up for different tasks. To really nail their point in to the thin air it was already riding on, they say that "there's not a single game that supports multi-threading even at a basic level."


My opinion, assembled quickly from a couple days of opinion-making:

The (well, a) point of multi-core processors is to not have to distribute these jobs to specific hardware, but to instead keep everything centralized under one type of hardware, with the specializing left up to software.
Hardware support is hell as it is. Adding yet another component to the mix would be a pain for everyone but the hardware sellers (who would be happily rolling in their money generated from false need, as they'll do).

On the other hand, keeping it to a single widely useful chunk of hardware, the system is easier to manage. Expandable and adaptable processors are the future, not yet more specialized equipment. The video card and sound card exist because they output to those types of external hardware, but a PPU does not have any magical role aside from being another processor. Therefore, what the heck is the point?

To add to it, I wouldn't mind PPUs as a choice (for those people who just go with what exists instead of waiting for Hell to freeze over) if they were not sold in such a way that, without them, you get arbitrarily locked out of features. There isn't a huge amount in the PPU scene, so the software development is very closely linked to the idea of selling hardware.
(With that in mind, actually: Is a PPU going to help anything but one specific physics engine? I doubt that physics engines work alike, so it seems unlikely that Ageia's PPU would be useful for much other than Ageia's physics engine).


Regardless of their claims, the fact remains that multicore is where computers are headed, PPUs or not!

By the logic they are using, we should also have:
-Game NPC hardware
-Rain drop processors
-Game networking processors

Oh, and those pesky "normal processors", whatever they do... (Processing or something?)

They're worthless, really.

doesn't matter either way since Physx can make use of multiple cores if you want. Or the PPU if you have their PPU processor. The PPU sounded impressive, far faster than a CPU with 10's of thousands or simple rigid body cubes being bashed about.

No ones going to make proper use of the PPU though, too pricey at $200 - $300. I plan to get a quad core CPU and run my game on one core, use ogres media streaming on a seperate thread to handle my scene manager, and use another core for Ageia Physics. Long term goals, but all the tech is there already, just have to have the hardware to use it.

The trouble is, yes, stream style processors are going to be so much more powerful at doing these types of computations than current x86 processors, but every new graphics card coming out from this point onwards is going to be a general stream processor! Just execute physics on that and be done with it. PhysX can compare their product against CPUs all they want - they'll win, based on faulty metrics.

Stream processors are going to be eventually included in normal CPUs (AMD's Fusion is one such project). The theory behind stream processors is basically that the returns on adding transistors to speed up one processing core (pipelining, unrolling adders etc) are increasingly small as you throw hundreds of millions more transistors at it. If you break away from that mould and write software (a renderer is the canonical example of this) that can efficiently run on multiple slower cores, then extra transistors can be used to add more cores, giving you a nice, almost linear boost. The computation power of an 8800 Ultra is an order of magnitude greater than the computation power of a core 2 quad, even though their transistor count is comparable (about 600 vs 650).

So yes, current processors suck for things like physics, but they'll hopefully transistion to the point where they can execute a small number of high performance threads and many lower performance threads simultaneously.

In the beginning i would have said yes. When looking forward to parallel designs like intel's larrabee i say no.

I used to think Physics hardware would be a great thing but with the rapid appearence of quad-code systems round the corner I dont think they are needed. My PC barely uses the two cores it has, but as Evak said, quad code, with each core doing its own thing, would be pretty awesome.

As long as its not as difficult to program as the Ps3 was said to be ;)


Hmmm, Just looked up Larrabee, read this news article: http://www.theinquirer.net/default.aspx?article=37548

BlitzMax needs threading...

Newton physics is utilizing dual cores also, no need to install additional programs like PhysX. And the Leadwerks 3D engine for BlitzMax is using Newton already, you can have easily over 100 moving physics objects with decent FPS even on the slowest dual core, which I have.

especially with quantum physics and other stuff on GPU, PPUs died before they really came to life ...

And I think this is just the start of GPU usage ... there will be definitely more to use them for that just the useless part they've done so far: eye candy with no content


PS: there is already an AI processor, so don't make fun of it, otherwise the next thing to buy might be an APU *lol lol* *sorry couldn't resist ^^*

yeah, were currently using Newton, mainly because its cross platform so we can use Blitzmax fully too.

I was playing with a GPU renderer for 3dsmax the other day, and my 7800GS ended up being 12 times faster than my A64 3000+. It's a real shame it doesn't support render to texture. It also overwrote the default cg.dll compiler that 3dsmax and ShaderFX use to create my shader programs.

If they get those things fixed it would be amazing :)

I think it could be a nice idea if it really was many many times more powerful than it is and did things that you just would not be able to do with a regular CPU, but from what I gather that's not so.

I guess I'll stick with tetris on a gameboy while waiting to get laid.

I think they would make most sense if built into consoles. They are a locked platform, so everyone who had the console would have the card. Of course, since this generation of consoles just rolled out the door, this is unlikely to happen. And by the time the next generation of consoles come around, we'll probably have some magic processors that can do procedurally animated characters swimming around in real time fluid dynamics liquids. :)

Not needed. At least not till games start taking advantage of them, and games will not start taking advantage of them till they are more standard, and they won't be more standard till games start taking more advantage of them...

The lesson to be learned here is simply, don't buy stock in Aegia. LOLZ...