One of the problems in the design work of phased arrays is the one taking into account phase shifter discreteness. Usually this problem is solved by means of finding the required continuous phase distribution and subsequent its rounding-off according to phase shifter discreteness. Such an approach is very simple but it often leads to non-optimal results, especially in the case of large shifter discrete-ness. Therefore, the problem arose to find the so-called discrete methods of phase synthesis that would allow to take into account the problem discrete nature more efficiently. From the mathematical point of view, these are the methods of finding the extremums of quite complicated functions in discrete space of search. Nowadays there is no the developed technique of such a task solving so it can be solved by different methods. In this paper it is proposed to use genetic algorithms (GA) to solve the problem of discrete phase synthesis because just these algorithms can easily be applied to the problems of optimization for complicated, multiextreme, nondifferentiable functions as well as for discrete search space [1]. The work is intended to show the advantages of the discrete phase synthesis method in comparison with classical method of continuous solution rounding-off and to investigate the possible ways of more rational computation when using genetic algorithm for phased array synthesis.