Poorly simulated processes due to insufficient resolution, such as cloud formation, precipitation, pose considerable uncertainties in both numerical weather prediction and climate-change projection. Exascale computing presents a promising future as well as challenges for earth system modeling, as the previously unresolvable processes can be explicitly simulated with unprecedented computing power. Globally cloud-resolving models (GCRMs) at the grid spacing of 3-km or finer have strong substantial scientific and societal impacts, like predicting extreme weather, simulating the hazardous natural events, even improving wind-turbine efficiency. However, to take full advantage of the E-class HPCs, geofluid dynamical models not only need to innovate advanced algorithms, but also adapt to modern computing architectures. In this work, we describe how the FV3 utilize the ultra-high grid resolution and build a dynamic-physics-unified self-contained framework. Exploring GCRMs with E-class HPCs is also a shared motivation among major modeling centers around the world, therefore results in the first intercomparison project of global storm-resolving models - DYAMOND. This work will also present FV3's participation in this project.