Your in a room with 20 people. Most of em are using 1% of their processor. Why can't those resources be shared and used where they need to be used?
Like the cpu usage is a giant pool, with 100%dominance given to the local system. But what's available is offered to the group. Sharing ram, cpu power, even graphic processing if all the systems are the same.
In my world, even with large airframe, one (good) system is enough. There is still that multi-threading bottle neck that NX is not achieving in most areas except faceted bodies. I'd much rather have Cavity Mill divide itself up amongst my 4 processor cores than the proposed generating of multiple operations in the background over multiple cores that's coming out in 7.0. I find myself mostly needing to generate an operation first to dictate how the next one will be dealt with.
For the real world, I'd look at CPU's with lots of cache. That's why the Xeons (12mb cache) are so popular for cadcam. Now, Vericut loves multi-processors.
Now a few years back I went to Blur Studios (SFX and movie animation house) with a friend. There they had render farms that ran 24/7 for all the frames of animation. Every computer in the place could go on line. So yes that is being done but for a specific purpose.
Well, I'm going to find out. They are setting my up with access to 3 computers in the design room. I'm on 32 bit, and I'm assuming the machines I link to will have to be 32 bit. Of course I'm not sharing resources, I'm just using them via remote desktop or comparable software. 3 virtual desktops all running UG, with my box left to just manage the desktops.