Abstract. Non-staggered, fractional step Navier-Stokes solvers are presented and analysed in comparison to a standard staggered solver. It is shown analytically that the non-staggered schemes have the same order of accuracy as the staggered scheme, although one of the schemes is non-elliptic in the pressure field at the grid scale. Numerical results are used to validate the analysis, with only the non-elliptic non-staggered scheme showing a reduced accuracy and efficiency compared to the staggered scheme. Details of a computer cluster built using Pentium 4 processors with a fast ethernet switch are presented. A serial benchmark demonstrates the excellent performance of the P4 for large array size compute intensive jobs, while both MPI and HPF coded parallel benchmarks show the effective parallel performance of the cluster.