Having listed the suitability of the newer style of parallel computers for neural modelling tasks, it is important to examine why they are not in wider use. Traditionally, parallel computers have been much harder to program than traditional computers and the typical neurobiologist was expected to understand a lot about advanced computer science issues before he or she could adequately construct neural models on such a machine. As this is not considered a reasonable expectation in neural modelling circles, most modellers have continued to use traditional (serial) computers and have had to sacrifice model detail in order to get acceptable performance figures. Such cut-down models may still require more than 12 hours to complete on a traditional high-performance computer [Bhalla:92a]. Parallel computers hold the promise of allowing more detailed models to be run in an acceptable period of simulation time. In order to make this power available to a range of neural modellers, it was decided to produce a version of a widely used neural simulation system [Wilson:89a] which could take advantage of such a parallel computer with only minimal manipulation by the neural modeller.