Super PI scores and overclocking

[goreblast]

Just doing some Super PI baseline benchmarks for a wee project I am thinking of doing. Should, provided all other variables such as other programs, cpu speed and ram timings are constant, the scores for each type of superPI be constant, ie if you run 1MB 10 times using -

defrag, restart, superPI run

I ask as for instance for the 1MB test after 10 runs I have
31.719
31.844
31.719
31.750
31.781
31.922
31.922
31.765
31.875
31.922

The average is 31.822 - should they not all be the same or very near it?

For the record the variation in 8MB over 5 runs is 13sec from highest to loowest and for 32MB over 5 runs it is almost 1minute!

[jrw]

Thats some big variations!

Shows why pi-benching demands perseverance.

Whats your project?

[goreblast]

I was going to try all the different tweaks to improve superPI performance across win2k, winxpsp2, winxp64 and win2003server. Testing 1MB, 8MB and 32MB. I was going to post results both here and on my own web space.

The scores above are a clean install with all high priority updates, all drivers except graphics drivers (cannot get crossfire to work in 2000).

Hardware - opteron 144 at fx55 speed with 2gb ram at 234 (3-3-2-5) on dfi rdx200 cf-rd

[The signature isnt the machine this is being run on]

[goreblast]

Anyone? I tried searching to see if these scores are consistent, what from everyones experience are consistent scores. Or do you need always to do thest say 5 times, take an average and use that as your "score" for comparative purposes.