Indirect effects of PPUs

[GoodRevrnd]

Other than the obvious effect of vastly increased physics effects, what indirect benefits do you think PPUs will afford? Do you think the freed up processing power will be better utilized for superior AI, or has AI theory really not advanced far enough to really reap the benefit of the freed up power?

Any other thoughts?

 

[SynthDude2001]

I don't know if I can see developers really utilizing processing power freed up by the PPU, for AI. They'd still have to support non-PPU-containing computers as a baseline...

 

[GoodRevrnd]

Well, we seem to have two working theories on the PPU field... On the graphics card, and independent card. What if on graphics card became the norm?

I guess we should probably accept the fact that according to Valves GPU survey (if I remember correctly) the average user is at ~ ATI 9600 technology, and we're in a day with the 7800GTX at the high end. While you can make graphics somewhat scalable, AI seems like it'd be absurdly tricky.

 

[Elcs]

Originally posted by: SynthDude2001
I don't know if I can see developers really utilizing processing power freed up by the PPU, for AI. They'd still have to support non-PPU-containing computers as a baseline...

Once (or if) the PPU tyhing takes off, they will end up excluding the people that dont have PPU's after a certain time. Then more specific optimisations will come in and increase performance/effect.

I dont think taking the Physics away from the CPU will do that much performance-wise. Physics implemented in current games probably takes up >5% of the CPU cycles in an average gaming session. Giving a CPU >5% extra power isnt much to play with.

Ill quote from Jeff7181 because we had a decent discussion about it in another thread.

Right... but a PPU doesn't allow better physics... it allows MORE physics. I think you're saying the same thing I am, but thinking of the application, not the technology behind it. All that stuff you said will all be possible because a PPU can do more physics calculations than a CPU can, not because it's more accurate or something like that. All that is possible right now, but it would be too slow because of the extra physics calculations required for the physics on all the extra objects.

*EDIT* A PPU isn't doing something a CPU can't... it's doing more things faster than a CPU can. The birth of the PPU is essentially the same as the birth of the GPU when you think about it. It's specialized hardware created to do a specialized task faster than a CPU can do it, which allows more work to be done in the same amount of time.

Apart from making a game with many more physics calculations to provide more realism to a game (dynamic wind, physics calculations for single droplets of rain, boulders rolling/bouncing/breaking up in realtime), I dont think the PPU will do much more for the industry instantly such as freeing up CPU cycles to improve AI calculations etc... but I do believe that it may change the focus of the industry into looking at the certain parts of games so overlooked these days.

The gaming industry seems focused on making pretty graphics at the moment, maybe introducing the PPU will make developers look at AI or other aspects of gaming.... You never know, they may take Red Factions Geomod technology and make it fit into gaming better. Imagine in BF2, digging your own trenches under enemy fire.... rushing from your trench to a nearby crater left by the enemies pre-emptive artillery barrage.

 

[phisrow]

The tricky thing about PPU adoption will be the fact that all the coolest uses of a PPU don't degrade cleanly in the case of a user without such hardware, which will make balance a real issue. With the GPU, things are much easier. You buy a better graphics card, you get better graphics. All else being equal, better graphics will be a slight advantage(particularly when sniping and so forth); but the effect won't be significant. Particularly by cutting polygon count and texture quality before draw distance and frame rate, you can keep everyone playing the same game without much trouble. But what do you do for the PPU? Either you make the physics effects pure tinsel(people with PPU have explosions calculated in realtime, the hoi polloi get canned effects) or you run into a problem: if the game depends on the physics effects in some way(deformable terrain, damage model based on Newtonian physics, etc, etc.) it'll be really hard to keep the users with real physics and the users with faked physics from diverging.

 

[TraumaRN]

And consider something else....PRICE....if they make the technology affordable and make people want it then you could(maybe) have a smoother transition enough people have to buy it for developers to start coding for it thats for sure...Myself I wouldnt pay more than 200 for something like that and the other problem I see is how does the market scale on such a product, it's obvious in the GPU market, but in a PPU market I'm not so sure besides making things faster.....*shrug* we'll see in Feb 06

 

[GoodRevrnd]

Think we'd ever see an AIPU? Or that such a thing even has any merit?

My concern is what phisrow wrote. If i'm investing in a $200+ card just for physics, I'd really like it to be game effecting, not just extra glitter. If that's the case, then they will almost certainly be relegated to a limited selection single player only games. And that's probably the best case scenario. =/

 

[GoodRevrnd]

Here's another point... if dual cores become commonplace how good a substituewould those be for physics processing? Do PPUs allow for really specific instructions based upon their architechture and optimizations, or is the point more to offload work from the CPU. I don't see them being quite as specialized as a GPU is (at least not initially).

 

[Mocib]

Originally posted by: GoodRevrnd
Think we'd ever see an AIPU?

http://www.aiseek.com
http://www.extremetech.com/article2/0,1697,1855163,00.asp