In this paper, parallel processing computer system models are considered, in which, each CPU service time of programs subdivides into independent parts (tasks) that can be processed in parallel.Different methods of partitioning a CPU service time into independent tasks are considered. Their effectiveness to system performance with relation to parallel processing with no multitasking in various types of computer systems and for different degrees of multiprogramming and variation coefficients of the CPU service times is investigated.Closed cyclic queueing network models of computer systems are considered where the CPU consists of two iden-tical processors connected in parallel. The results are obtainedusing simulation techniques.