A parallel processing system is described that consists of a minicomputer host and a set of bipolar microcomputer modules. It is argued that such a system in which the microcomputers operate with little mutual interaction should be effective for an important class of problems in numerical analysis. In particular, estimates are given for the operation of the system on a problem in gaming theory. In this problem, the extensive I/O and software capabilities of the minicomputer provide ease of use for a large part of the problem. The relatively simple part of the problem, which requires almost all of the computational time, is executed in parallel on the microcomputers. It is argued that the system, with 10 to 20 modules, would offer one to two orders of magnitude more speed at several orders of magnitude less cost than current large general-purpose machines. The potential for the development of new algorithms that exploit fully the characteristics of the new devices is discussed.