We present a quantitative method for selecting optimum configurations of computer systems. For each configuration we specify a set of requirements reflecting user's needs. The level of satisfaction of requirements is called the global preference score. Using this indicator and the total cost of each configuration we solve the following optimization problems: (1) select a configuration which maximizes the global preference score for a constrained cost of computer system, (2) select a configuration which attains a given level of global preference for a minimum cost, and (3) select a configuration which attains the maximum global preference/cost ratio. We first present the optimization method and then a complete case study of computer optimization.