@nox once optimized no difference.
i suppose stagger size equal to number of nonces would be more accurate. the idea here is to have one long continuous read of the scoop data by the miner and that is what optimizing provides.
again to the referenced post
@Blago's plot experiment
17930413153828766298_69900000_1000000_4000 - not optimized, reading time 3818 ms
17930413153828766298_64900000_1000000_1000000 - optimized, reading time 2143 ms
the second to last number in the filenames is the number of nonces and the last is the stagger size. optimized they are equal.