Astrometric microlensing is a unique tool to measure stellar masses. It allows us to determine the mass of the lensing star with an accuracy of a few percent. In this paper, we update, extend, and refine our predictions of astrometric-microlensing events based on Gaia's early Data release 3 (EDR3). We selected about 500.000 high-proper-motion stars from Gaia EDR3 with {mu}tot>100mas/yr and searched for background sources close to their paths. We applied various selection criteria and cuts in order to exclude spurious sources and co-moving stars. By forecasting the future positions of lens and source, we determined epoch of and angular separation at closest approach, and determined an expected positional shift and magnification. Using Gaia EDR3, we predict 1758 new microlensing events with expected shifts larger than 0.1mas between the epochs J2010.5 and mid J2066.0. Further, we provide more precise information on the angular separation at closest approach for 3084 previously predicted events. This helps to select better targets for observations, especially for events that occur within the next decade. Our search lead to the new prediction of an interesting astrometric-microlensing event by the white dwarf Gaia EDR3-4053455379420641152. In 2025 it will pass by a G=20.25mag star, which will lead to a positional shift of the major image of {delta}{theta}_+_=1.2_-0.5_^+2.0^mas. Since the background source is only {Delta}G=2.45mag fainter than the lens, also the shift of the combined center of light will be measurable, especially using a near-infrared filter, where the background star is brighter than the lens ({Delta}Ks=-1.1mag).