Think of your mesh and relative cell size as thousands of cross sections. Each cell (up to 8 faces) has multiple property tables and is a cross section. The d(t) is also important too. If you are getting realistic solutions to a model with a high cell size and large time step than it will be efficient. The more cell refinements, the more 1D/2D connections the lower your time step, the warm up period, all of these things impact run times. Every situation is different
Luis, thank you for your thoughtful response. I guess to be more specific, I am wondering how HEC-RAS handles the dry cells in a mesh. If a majority of model runs will generally have flow limited to the main channel (within the levees), and flood plains beyond the levee will generally be dry unless modeling > Q25, does it make a difference if the model is split into separate 2D areas?
Will the model run faster and more efficiently in a low flow situation when the wetted cells are limited to the main channel and there are three separate 2D ares/meshes (i.e. one mesh for the main channel and two for either side outside of the levees) or do dry cell not affect the model run time?
I would recommend doing it as a single 2D area. The connections between two different 2D areas can be a source of instability and slow computations. A single 2D area will be more stable.
Having a large amount of dry cells can cause an increase in model run time, but it is usually not that much. I believe that RAS has built in logic to minimize the additional model run time from dry cells.