In a Sample n function, the amps will choose rows in some random fashion. So there is no need to sort the rows.
In TOP n function the rows are sorted first and then the top n rows specified are returnd.
You can use random if you want to get a truely random result.
SAMPLE (RANDOMIZED ALLOCATION) returns a truely random result while TOP n (without PERCENT/ORDER BY) simply returns the first n rows found on a single AMP (or multiple AMPs). SAMPLE is always more overhead/slower compared to TOP (without PERCENT/ORDER BY)
If you just want to see some data better use TOP, if you do some statistical stuff switch to SAMPLE.
Btw, what is MECONOSAM LEVEL?