The architecture and performance of a multicomputer-type digital signal-processing (DSP) system are discussed. The NOVI system has been created to examine methods for organizing parallel DSP systems and developing parallel programs for a wide range of DSP applications. NOVI presently consists of 36 processing nodes, in each of which an INMOS Transputer is used as a CPU. Its parallel program debugging assistant system facilitates powerful debugging functions to observe all program states. A load-balancing technique using interframe prediction and layered large-grain data flow is proposed to achieve higher performance for data-dependent algorithms such as motion-picture coding. The expected performance of the NOVI system with the proposed technique is also discussed.