In this paper the parallelization of a evolutionary neural network optimizer, ENZO, is presented, that runs efficiently on a workstation-cluster as a batch program with low priority, as usually required for long running processes. Depending on the network size an evolutionary optimization can take up to several days or weeks, where the overall time required depends heavily on the machine load. To overcome this problem and to speed up the evolution process we parallelized ENZO based on PVM to run efficiently on a workstation-cluster using a variant of dynamic load balancing to make efficient use of the resources. The parallel version surpasses other algorithms, e.g., Pruning, already for small to medium benchmarks, with regard to performance and overall running time.