I read an article in EE Times this month that said the software really can't exploit those multi processors all that well and that you are really not getting the speed you are paying for.
I am only running single core machines anyway but I don't do much that is speed sensitive.
I do believe most of this quest for speed is driven my the bloatware Microsoft sells. Most of the time I suspect slowness is from "paging" in more code when you run out of memory. XP seems to have about a half gig of code loaded before you actually start doing anything. I do try to keep my active tasks (ctl alt delete screen) down to a page.
If you run W/98 on even a 1gz machine with 128-256m of memory it is a racehorse. The junkbox video server I have back in my computer room can stream a Replay TV show and still do some work (copying disks or something). The bottle neck seems to be the disk controller.
I thought about throwing a SATA controller in there and using the IDE for one thing and the SATA for another to see if it makes a difference. It could also be the actual memory fetch. I am not sure how many paths to memory they use these days. I lost track of that kind of detail when it all went to one chip.
I still miss CPUs with 1000 cards in them ... a little
