This work compares the predictive performance of RANS and LES solver in capturing the effect of terrain and wakes on the performance of the Bessaker wind farm. This 25 turbine wind farm is located in a highly complex terrain and is exposed to predominantly high westerly and south easterly winds. A one-equation sub-grid scale LES turbulence model has been used to help capture the wake dynamics: particularly the effects of wake meandering and wake-turbine interactions. A comparison between RANS and LES models highlights the influence of turbulence model on wake decay and its subsequent effect on prediction of power production. The LES model predicts delayed decay of the wake and more pronounced wake interference leading to a lower power production in wind farm than the RANS case. The RANS model overpredicts turbulence, which cause faster turbulent momentum diffusivity and faster wake recovery. This study has given some insights regarding the power production at Bessaker wind farm for neutral conditions and westerly flow.