The problem with PPUs seems to be that, while they can calculate tens of thousands of physics objects, are today’s GPUs actually capable of rendering that many objects at a decent frame rate? From what’s been reported about GRAW, I would say no.
Perhaps PPUs might be useful for online gaming where the physics have to be calculated by the server hosting the game. Ie you’d have a dedicated server with a PPU effortlessly handling the physics for all the players connected.
IRT thewonderfulcar
Good idea about the servers, but then again, couldn’t the idle GPU be put to good use?
In the end it’s all Floating Points, so it could probably take some work from the CPU.
More likely the physics module of the Doom 3 engine would have to be rewritten to utilize the features in the PhysX SDK. In any case why change the physics to accomodate a likely novel technology? I like my trickjumping as it is and prefer that “real” or “fancy” physics not to interfere with it. The PPU is an as yet proven technology and until it is proven to be a necessity to the alternative of adding to the clock cycle time of the GPU, it is not needed.
Yes, and both of which are not out yet. :)[/quote]
Cell Factor is out…
http://ageia.com/physx_in_action/cellfactor.html
[/quote]
No… Cell Factor DEMO is out. Demo is not the full game…
Cellfactor + physX = Drool factor
Check out the cell factor HD video on Aegia’s site
http://www.ageia.com/physx_in_action/footage_physx.html
it is amazing stuff, but it’s frivalous when compared to gameplay and of course …this + great gameplay = wow.:faint:.
some more thoughs
Perhaps PPUs might be useful for online gaming where the physics have to be calculated by the server hosting the game. Ie you’d have a dedicated server with a PPU effortlessly handling the physics for all the players connected.
maby but how much extra data would have to be esent over the intent. your realy going to limit the use of physics that way and would be little point putting such an extensive physics system in the game
by looking at CellFactor i made two judgemnts. Small level and small number of players. If there are a small number of players then the server can give every clint exactly what every player is doing at every frame, right? so the physics could be calcualted after every player recives on sever frames worth of data. all player will have axactly the same data frme wise and the card will process this exactly the same same way for everyone. the end result they have the identical resulting frame. so no additiional data is need for the actul phsics
It’s easy to parallellizise gfx because you send instructions to it and forget it until next frame while with physics you send calculations and then have to poll for the results. Requiering some form of threading or similar solution.
maby not. graphics is mainly sending data to the card. would physics card not already hold the data on the crad all the time then send mainly back to the CPU rather then the CPU only sending small amounts of data e.g. where a rocket has exploded. and there is little need for CPU control like someone said you cant optimise physics buy cutting out unseaable calcualtions
In a few years anyone will have a 100MBit/s connection - and then you can render everything on the server and stream it as video to the client.
And think of the consequences: No more cheating, no 500+ bucks GFX cards or expensive CPUs … 
considering this is looking at the near future and PPU then its a bit OT imo
In a few years anyone will have a 100MBit/s connection - and then you can render everything on the server and stream it as video to the client.
And think of the consequences: No more cheating, no 500+ bucks GFX cards or expensive CPUs …
As far as I could gather (and don’t assume I’m correct), Direct3D10 (aka DirectX10) features massive alterations in DirectX technology that it will require a new series of compatitible cards.
I mention Direct3D10 as physics etc feature heavily within it’s new capabilities. The PhysX card looks pointless imo.
For info, Direct3D10 requires Windows Vista and a compatible graphics card (not released afaik??).
Btw, Hi!!! 1st post for me.
No… DirectX10 has nothing to do with anything specifically with physics…
But it’s no secret that Microsoft seems to be coming up with a DirectX physics (DirectPhysics?) thing… Which is not a bad thing IMO.
The technology name is DirectX. Direct3D is just part of the api - like DirectDraw, DirectPlay, DirectSound, DirectInput, etc.
I think the future is multiple GPUs, then one of them can be temporarily assigned to some physics.
Not for quite some time. Current GPU’s are just not designed for the sort of calculations ‘real’ physics require.
I don’t know about anyone else, but I’d rather spend that $300 extra dollars on better memory or video card.
Anyone see any stats on how many physx cards have been sold?
You don’t need a PhysX card to run cell factor, there is a command line to add to the shortcut of the demo: EnablePhysX=false. I’ve tried it and it runs really well even on maximum settings withouth a PhysX card.
I’d hold onto your cash. 
Z
You don’t need a PhysX card to run cell factor, there is a command line to add to the shortcut of the demo: EnablePhysX=false. I’ve tried it and it runs really well even on maximum settings withouth a PhysX card.
I’d hold onto your cash.
Z[/quote]
Stuff like the cloth simulations are not performed in software mode. If thye were enabled, I would assure you that your fps would suck ass
hmmm,
Unfortunately, even in the gigantic PhysX tech demo we know as Cell Factor, we can see the limitations of PhysX’s current system. First off, in order to run the demo smoothly, you have to tone down your graphics card to a fairly low resolution. Even with a GeForce 7900 GTX 512 MB card, the system would slow down during intensive situations. Even at low resolutions, the system can be brought to a halt during activities which require lots of physics calculations. This can be seen clearly in the last video above.
Other games which have early PhysX support are also stating that this is the case. Whether this is a limitation of the PhysX software API, the game engines at play, the limitation of the PCI bus, or an inherent flaw in the PhysX PPU. The fact of the matter is with the first batch of PhysX titles on the market, the PhysX card is slowing down systems rather than truly accelerating physics content. While PhysX enabled titles have bigger, more dramatic effects and more debris, this is at a cost of slightly slower performance. Now, if one is spending all sorts of cash on a high-end physics accelerator one would expect the title to run with bigger effects and run smoother, not one or the other. It’s likely that these early quirks will be smoothed out in time, but early adopters are (understandably) not thrilled with the performance trade-off they are having to make thus far.
Both ATI and NVIDIA are considering adding more physics capabilites to their next generation GPUs.
You really think they are going to let one small company take a bite out of their gaming shares!?
They’ll try anything in their path (as long as it’s legal)
Yeah but they’re doing it via DX 10 Shader Model 4.0 which actually reduces graphics load by rendering differently and using the GPU entirely for graphics calculations rather than the GPU / CPU in combination.
Expect DX10 SM4.0 to offer big jumps forwards in both quality and the number of renderable objects whilst maintaining or even increasing FPS>
Al.
yah, maybe, it’s all fine in theory and until someone actually makes a DX10 SM4.0 vid card no one will know