Anyone here know much about PPUs? They seem to be the next big thing for processing... But are they really necessary? There are no games out for it yet, and no major companies seem to be embracing it. What do you think about the PPU, and its questionable future? I mean, will we HAVE to have one to play games in 2007, or will we be able to play like we do now? Will some games require them and others not?
This is very hush-hush, but they CAN be purchased before the set date (in May) through some sketchy-looking sites. The dudes on [H]ard OCP said that http://www.overclockers.co.uk/acatalog/PhysX_Accelerators.html actually will sell them to you, but I am not sure if they ship to the USA. At the moment though, there is no point playing it, unless you want to play the demo, which you can play without the PPU
It's needed, physics on the CPU is horribly inefficient... There is one game that is supposed to be awesome with a Physx unit, i forget what it's called, but the entire game is a first person shooter based on gravity.
A lot of games don't seem to have any "weight" to them when you play, half-life 2 was only a taste of what can be done, and I hope that these ppu's become mainstream/cheap/ and integrated onto GPUs...
Here's the site of AGEIA, the maker of physx. You can download there some presentations, movies, screens. www.ageia.com
The PhysX processor is very powerful. Normal CPUs can calculate the phyics for maybe 50 objects on the scene without slowing the animation. PhysX can calculate physics for over 50 000 objects without slowing animation. PhysX also calculates the phisycs better, more realistic, like in a real world. In game which supports physics everything will be interactive, you will always be able to crash the wall with bazooka. The systems like Havos just simulates the physics and not for all objects so you cant destroy the wall in Half-Life 2 with bazooka. Look there: http://physx.ageia.com/footage.html there are 3 real-time footages. Looks amazing
PhysX is not the only thing to calculate physics. There is also Havok FX. Havok FX is not a card, it's technology invented by Havok and it will be used with nVidia cards. You can use Havok FX if you have GeForce 6 or GF 7, but it's recomended to have the stronger cards in SLI to calclate the physics with good speed. To use havok fx you will probably need to have some havok software. I don't like Havok FX because:
1. It uses GPUs performance (one pipeline if I remember good)
2. It's not as good as PhysX (slower)
I read about this a while ago in some magazine, i think it was either CPU or PC Modder. But they had an article that these things were in development set for release in either late 06' or early 07'.
Pesonally, being a late adapter of new hardware, and an early adapter of new software...ill probably have to make an exception on this thing. For example i JUST recently bought a FX5500 and have OCed it to be pretty close to some comparable new cards out on the market today. But that just goes to show how late im talking about when i say im a late adapter. However, since new games and software may require this as early as the fall of 07', it is likely that i will have to be an early adapter of the PPU in order to be an early adapter of the new software.
But image if they some how could figure out how to put utilize this hardware for an OS, i cant image what they could do with it in that sense, but im sure that there are PLENTY of possibilities that microsoft and linux developers are already playing with.
While the video and the idea of a PPU is impressive, games that will use the PPU in the near future will ONLY use it for effects. Otherwise, non-PPU users will not be able to play them. The future of this card has so much potential, but the main thing I am wondering about is how it will evolve... Will it be like a vid card, where there are better and worse ones? Or will it be like a network card, where every one performs about the same. The possibilities with everything are endless.
And Sebaci, the CPU of a computer can do around 500 (with 1000 or so on a dual-core) object calculations. The PPU will start being able to do around 5000. So there is a 10x jump there. Not quite as significant as you were suggesting, but still there
As for the Havok engine, it will run mainly on SLI systems (OR crossfire ATI systems -- ATI will allow the cards to be different though, so anyone want to use an old card as a surrogate PPU? ). It will completely ignore the second GPU in terms of SLI, only using it for physics processing. Therefore, it will probably use ALL the pipelines. But either way, it will still be MUCH slower than a direct PPU, only processing 2000 or so instances at a time.
Another thought on all of this... Is it really necessary? No. And I will explain why. As of now, all calculations on a figure are performed like the objects in the HL2 engine, because it is the most revolutionary. Being a mapper for CS and CSS, I know much about the engine. Calculations are performed VERY VERY inefficiently, with an object being split into basic polygons, with the forces then simulated on each individual poly, which adds to a whole vector for response movement in the 3d virtual world. If another engine came along (HL3 plz?) that would revolutionize gaming like HL2 did, then it would likely learn to do these same calculations on one figure as a whole, modifying the verticies and being able to calculate and recalculate the whole figure, instead of its parts. "But what if it split into pieces, squirrely?" Well, it would do the calculations independantly, JUST LIKE THE PPU. Unfortunately, the whole prospect of this revolutionary engine is extremely difficult. Advanced physics calculations on a scale like this have never been attempted, and hopefully, the geniuses at Valve are working on something of the sort now.
But whatever happens with all of this... We will be blown away in the near future by insane effects, mind-blowing graphics, and almost real-life physics on games. I look forward to it.
|The PPU will start being able to do around 5000 |
Much more, really much more.
|As for the Havok engine, it will run mainly on SLI systems (OR crossfire ATI systems -- ATI will allow the cards to be different though, so anyone want to use an old card as a surrogate PPU? Smile ). It will completely ignore the second GPU in terms of SLI, only using it for physics processing. Therefore, it will probably use ALL the pipelines. But either way, it will still be MUCH slower than a direct PPU, only processing 2000 or so instances at a time. |
Where did you read that? Havok is not the same as Havok FX. Havok FX will run only with nVidia cards, ATI has nothing to do with Havok corporation, and ATI works on his own integrated physics accelator in radeons
I think PPU is necessary. Havok in Half-Life 2 is weak, has bugs and not all objects are interactive. As I said, you can't even crush the wall with bazooka
Not that it matters anyways (but I would still like to know), where did you see that it would display more than 5000 instances? The dudes on [H]ard OCP (that actually have the card) say it is 5000. I can get a link if you really care
I read about the Havok also on [H]ard OCP, and I did re-read my post, and I did imply that Havok was going onto ATI cards. Like you said, this is completely false; they are working independantly. I meant Havok FX when I talked about it in my last post... It seems almost common sense that I would not be talking about a preexisting engine.
While the current physics are weak, what you are suggesting is probably far in the future, when PPUs are on every machine. PPUs in the near future will simply be doing effects and such, because the games that will use a PPU will have the PPU use optional.