Summary: The use of a high-order discontinuous Galerkin/interior penalty method for the Reynolds’ averaged Navier-Stokes computation, based on the Spalart-Allmaras model, of compressible turbulent flows is investigated. In particular different stabilizations of the iterative process are compared and discussed, including a new definition for the penalty parameter. Based on grid convergence analysis for the flat plate boundary layer, as a function of interpolation order, element type (triangle, quadrangle) on the one hand and grid spacing and stretching on the other, clear guidelines on the choice for boundary layer resolution for practical applications are provided.