ok, so I didn't read the entire thread, but I did read about half of it.
I was thinking about something.
if your game is running at 60fps, and the physx is also processing physics for each frame, as frame rate increased, wouldn't that require the physx processor to also process the physics faster? Just a thought, I wasn't really siding with anybody on the whole argument.
Originally Posted by Johnny Doe
PhysX is an extra, differental HW rendering of effects such as cloth, smoke and such. It's different from graphics rendering when rendered externally. That's what the original Ageia card did. It rendered PhysX near your graphics. Which is what dedicating a card does right now.
You don't seem to understand cause PhysX can't be rendered on CPU due the way it's coded. CPU is too slow/shader lacking to do it. A dedicated GPU renders PhysX whether you dedicate it or not. Even one card out of an SLi config is taken since it PhysX can't be rendered on SLi. Go back to the first page then look at the results between 260 SLi PhysX and 260 + 260 PhysX and so on.
I would just like to say, you are wrong about physx not being able to be processed on the CPU, it can be done. The reason why it is so slow is because nvidia purposing used a old slow language for the CPU software, instead of using a newer language such as the SSE. It is known fact, if you do the research, that if nvidia coded physx in SSE, or any other language that the CPU is efficient at, it would perform just as well as a GPU.
EDIT 2: if you are wondering, here is my source for that slander of info.
http://arstechnica.com/gaming/news/2...pite-intel.arsEdited by thrasherht - 11/1/11 at 5:02pm