Simulating the ocean’s submesoscale is key to understand the mass and energy cycles of the ocean and the global climate system. Contrast to the ocean’s mesoscale, submesoscale processes are usually highly ageostrophic and manifest at the scales within 10km. Ocean general circulation models with kilometer-resolution are capable to resolve key submesoscale processes, hence indispensable for both process and climate studies. We construct a grid hierarchy for the ocean-sea ice model in the High-Resolution Earth System Model on Sunway supercomputer (SW-HRESM), which is based on Community Earth System Model (CESM2) with deep optimizations on the Chinese home-brew supercomputing architecture of Sunway. The highest grid resolution is 0.03o (2.4km globally). In this study we evaluate the ocean-sea ice coupled simulations by SW-HRESM, focusing on the submesoscale and the kinetic energy (KE) cycles. In particular, highly ageostrophic submesoscale turbulence is simulated, dominated by deepened mixed layers (ML) during winter and the ensuing instabilities. KE and its transition between scales are further evaluated for major western boundary current systems. During winter, submesoscale is shown to dominate inverse cascading which energizes large-scale flows, as well as forward cascading to dissipation scales. The mesoscale-submesoscale continuum and the associated inverse KE cascading is further complemented by the forward KE cascading from the large-scale due to flow instabilities. In order to fully resolve the submesoscale spectrum, including frontal processes and wind-wave interactions, models finer than 1km are needed. Besides the model resolution, improvements for both the ocean and the fully coupled model of SW-HRESM are also discussed.