Physx Performance in Games

[phaxmohdem]

I've been goofing around with the Physx demos from nVidia today, and wondered what the ideal setup would be for actual games that employ the Physx engine (if in the future I ever get a title that uses it)

I currently have an 8800GTX driving my primary 24" LCD, and in my nVidia control panel I have my secondary video card (8600GT 256MB, which I use as a 2D card basically for extra displays) set as the Physx processor card. In the demos however, (particularly the screen saver demo) I noticed that it was much faster to do the Physx on the 8800GTX card that was also rendering the scene, rather than dedicate the 8800gtx to 3D rendering, and offload the physx calculations to the unused 8600GT card (Granted the 3d objects/rendering in the demo were pretty weak, and probably wouldn't tax a Geforce 256 let alone the 8800GTX)

I guess what I'm asking is, in real game scenarios, where the 8800GTX would be hard at work rendering the scenes, would it be of benefit to have the 8600GT card handling the Physx work, or just the big boy do both?

 

[Spike]

Hopefully others can answer with more information but I remember reading several posts by members who had done the same test as you and found little difference in having a second card as the dedicated physx processor. I'm wondering if this has more to do with the current physics effects being rather "weak" than anything else. If developers go all out with physX, as in make a game that can't really be played without it, I'm wondering if a dedicated physX card might have more of an influence there.

Sorry I don't have much more to offer, that is only what I have read.