The cpu entries for /proc/stat are: User, User1, System, Idle (cumulative)

Therefore for a delta T,

delta(User)+delta(User1)+delta(System)+delta(Idle) should have a constant value at delta T.

But this is not the case - at times when system load changes, the sum can shoot up and down from "100%".

Are the values recorded just not accurate enough at times of rapid change - or is there somewhere a set of more self-consistent data values ?

No - I don't want to install another package :-)

2-nov-05
Oh well, it was worth asking, even if no-one knew the answer ...