This study provides a novel contribution towards the establishment of a new high–fidelity simulation–based design methodology for stall–regulated horizontal axis wind turbines. The aerodynamic design of these machines is complex, due to the difficulty of reliably predicting stall onset and post–stall characteristics. Low–fidelity design methods, widely used in industry, are computationally efficient, but are often affected by significant uncertainty. Conversely, Navier–Stokes CFD can reduce such uncertainty, resulting in lower development costs by reducing the need of field testing of designs not fit for purpose. Here, the compressible CFD research code COSA is used to assess the performance of two alternative designs of a 13–meter stall–regulated rotor over a wide range of operating conditions. Validation of the numerical methodology is based on thorough comparisons of novel simulations and measured data of the NREL Phase VI turbine rotor, and one of the two industrial rotor designs. An excellent agreement is found in all cases. All simulations of the two industrial rotors are time–dependent, to capture the unsteadiness associated with stall which occurs at most wind speeds. The two designs are cross-compared, with emphasis on the different stall patterns resulting from particular design choices. The key novelty of this work is the CFD–based assessment of the correlation among turbine power, blade aerodynamics, and blade design variables (airfoil geometry, blade planform and twist) over most operational wind speeds.