Résumé
We describe recent work focused towards a better understanding of red supergiant stars using 3D radiative-hydrodynamics (RHD) simulations with CO5BOLD. A small number of simulations now exist that span up to seven years of stellar time, at various numerical resolutions. Our discussion concentrates on interferometric and spectroscopic observations. We point out a number of problems, in particular the line depth and line width that are not well reproduced by simulations. The most recent introduction of a non-grey treatment of the radiation field dramatically improved the match with observations, without solving all difficulties. We also review the newly revived effective temperature scale controversy, and argue that it will only be solved using 3D RHD models.