Computational simulation is now an essential methodology of science, along with theory and observation. The ability of scientists to understand and predict planetary climate variability largely depends on the veracity of the climate simulations produced by numerical models of the interacting components of the Earth system. Oceanic and atmospheric models are numerical approximations to continuous forms of the equations governing fluid flow and are “closed” by sub-grid-scale parameterizations that represent physical processes on temporal and spatial scales that are not resolved by the chosen model grid. In the past two decades, the rate at which the world’s fastest computers perform floating point operations (FLOPS) has increased by a factor of 10,000. This increase in computing capability has been exploited in several ways. Longer integrations for applications such as paleoclimate (Dijkstra and Ghil, 2005) and the inclusion into models of additional processes such as biogeochemical cycles and ecosystem dynamics (Moore et al., 2004), are two such examples. Another example, and the focus of the present study, is to increase the spatial resolution of models such that a greater fraction of the physical processes are explicitly resolved, and fewer are parameterized.