Nobody claimed it was. It's a test of how different machines run the same piece of code, which can be useful but probably isn't as useful as some people out there believe. If you're interested in how fast a machine can compute pi it might be a useful test though.
Isnt the real performance result VERY dependent on operating system and how it handles multiple threads and utilizes the cache?
Of course, and it's also dependent on the compiler, the algorithm, the amount of available memory, what other processes are running at the time, and probably the weather and what you had for dinner last night.
Linux results are not directly compared to Windows results since the Linux version just runs faster.
I used to download all kinds of benchmark tests. Magazines etc. often use them but good benchmarking is hard, and you have to specify what aspect of performance is being tested.
Some people use standardized demos in games to measure the FPS in different situations.
Some people use Linux/BSD/whatever kernel compile times.
My favourite screensaver makes my computer sweat much harder than the pi test. Seems silly to run a screensaver which makes the computer work harder than it does in general usage.
Nope. Or rather, I haven't tried it yet. I used to fold for the myelin project but now that I pay for electricity I don't leave spare machines on all over the place anymore.
I know the P4 mobo has on-board VGA... dunno about the Celeron board since I didn't bother to look. Besides, I will be using an ATI 1650x 512MB video card. I'm going with the P4 and it's accompanying VIA mobo. The BIOS is pretty decent and I can safely overclock it to 2.2GHz It's just a project PC I feel like fiddling around with... it should make for a decent back-up if I need one in a pinch.