Before you begin, look at the creation steps for the two methods from a “big picture” point-of-view.
Derive in-place part method:
When the command finishes, the part derived from the assembly is the active representation. The substitute level of detail representation is identified in the browser with the icon.
Part-on-disk method:
From an authoring point of view, the derive in-place part method is quicker most of the time. However, from a memory savings point of view, the simplified part provides more benefit since you have thorough and explicit control of geometric complexity.