But not all AML subtypes express CD34. So in these cases you need to sort them based on other markers expression (usually it might be CD117 - but more problematic would be cases of AML-M5).
Flowcytomertry sorting can provided more purity of cell by using multi-marker sort. Usually magnetic beads sort is operated easily. You have your choice between the two methods depended on your main purpose of your further study after sorting.
I agree for AML-M5 (monoblastic). CD64, CD11b and lysozyme are characteristically positive in AMoL. CD33 and CD64 are consistently expressed in 100% and 96% of vcases, respectively (CD33 is generally bright).