Simulations with the Community Atmosphere Model version 5 (CAM5) are used to analyze the sensitivity of the large-scale circulation to changes in parameterizations of orographic surface drag and vertical diffusion. Many GCMs and NWP models use enhanced turbulent mixing in stable conditions to improve simulations, while CAM5 cuts off all turbulence at high stabilities and instead employs a strong orographic surface stress parameterization, known as turbulent mountain stress (TMS). TMS completely dominates the surface stress over land and reduces the near-surface wind speeds compared to simulations without TMS. It is found that TMS is generally beneficial for the large-scale circulation as it improves zonal wind speeds, Arctic sea level pressure and zonal anomalies of the 500-hPa stream function, compared to ERA-Interim. It also alleviates atmospheric blocking frequency biases in the Northern Hemisphere. Using a scheme that instead allows for a modest increase of turbulent diffusion at higher stabilities only in the planetary boundary layer (PBL) appears to in some aspects have a similar, although much smaller, beneficial effect as TMS. Enhanced mixing throughout the atmospheric column, however, degrades the CAM5 simulation. Evaluating the simulations in comparison with detailed measurements at two locations reveals that TMS is detrimental for the PBL at the flat grassland ARM Southern Great Plains site, giving too strong wind turning and too deep PBLs. At the Sodankyla forest site, the effect of TMS is smaller due to the larger local vegetation roughness. At both sites, all simulations substantially overestimate the boundary layer ageostrophic flow.