trustworthy and reliable benchmark program?

[Burn]

Each time i use 3Dmark 05 or aquamark i have different scores every time.... for exemple if i run 3Dmarks 3 time in a row i could get a score someting like 6610 , second 6468 and third 6598... how i can effectively judge if a little tweak have his effect ??


anyone know a bench program with less errors ?? i was playing last nignt with Vid card PCI latency timming and lowered from 255 to 64 .. and raise it by increment of 64 ..... best performance seems to be between 128 and 192 but i can't pinpoint it because of benchmark innacurancy

[TwoRails]

I wouldn't consider just a few percent difference a flaw or unreliable. There are too many variables to be 100% consistent. Even very minor, and acceptable voltage variations from a quality power supply can affect such benchmarks. Little things like Window's services polling and grabbing a few CPU ticks will changes things. NICs with an always on connection can alter from 100% consistency, and so on...

[Burn]

Thanks for the reply. I admit that your are right. I was paying attention to the process and services running to be the same each time i lunch a bench program. I guess i can't tweak it more than that ....i shlould stop trying to make it faster and play games !

Everyone here know this : If it aint broke .. dont fix it.

well even after 13 years with computers i can't get used to it and sometimes it happened that the '' fixing '' attempt was a total disaster ... it is simply more powerful than me .. haha