>but i guess in order for all of it to work several hurdles
>would have to be overcome, this seems to have not been to big
> a problem for video ( sli technology ) so why would it be
>for the cpu?
For 3D graphics, its easy. Rendering is highly parallel by its very nature, its not a problem to let one videocard render half the screen, and the other the other half. (In fact, this has been done by 3DFX (Voodoo2) ages ago. in fact bis, if you look at a modern GPU, its extremely parallel itselve). Same goes for data storage (RAID is a similar concept).
Now, for game engines (and many other apps), its either hard, or sometimes impossible to spread the workload over different CPU's in an effective manner.
Consider this: say your workload is to mail 2000 letters. If that takes you one day (put them in enveloppes, add stamp,write address..), 2 people will do it in half a day, and 12 people in an hour. This is 3D graphics.
Now consider you have to <i>write</i> the letter, and it takes you a day. How long will it take 12 people do you think ? Are you going to let everyone write a single paragraph ? Thats gonna be one weird letter...
In theory there should be some tasks in a game engine that can be threaded, AI, physics and 3D modelling come to mind. In practice, apparently there is so much need for communication between those threads, that the benefit just ain't there. At least not yet.
= The views stated herein are my personal views, and not necessarily the views of my wife. =