It was proposed to investigate small-scale spatial and fast temporal scales with a threedimensional, magnetosphere-ionosphere coupling code. This required computational resources far beyond what a personal computer or work station could provide. Runs had to be made on the fastest (parallel, distributed memory) machines. An attempt was made at compiler-driven, "automatic parallelization"—but ultimately the code was explicitly parallelized. A 40-fold increase in performance was seen from going to the supercomputer (over the PC), and a further 40-fold speed-up was achieved from the parallelization (with 50 processors). This will enable further study of micro-scale physics in magnetophere-ionsphere coupling.