Plain Language Summary Many recent earthquakes only rupture in deep (>15 km depth) with the shallow portion of the fault unbroken in a longer time, including the case of the 2015 M-w 7.8 Gorkha Nepal earthquake in the central Himalayan arc. A few historical earthquakes seem also rupture only the deep portion, with much stress unreleased in the shallow portion as a stress reservoir. It is unclear how these phenomena are related to fault friction and geometry properties. We used numerical models to study the influences of these properties on stress accumulation during the interseismic period and on the time of final coseismic break. We found that comparing to shallow-dipping fault, steep-dipping fault can cause faster stress accumulation but is capable of sustaining lower stress in total, so it tends to break in a shorter time. On other hand, stronger frictional strength allows more stress accumulation in a longer time before the earthquake. In this way, the shallow portion of a fault with low dip angle and high frictional strength can be very stable in a long time and even absorbs stresses from earthquakes of the deep portion. So the shallow portion of this kind of fault is capable of hosting a very large earthquake.