Two widely used sub-grid scale models namely: the standard and the dynamic Smagorinsky models were tested in a simulation of the flow in a thermally driven 3D cavity at turbulent Rayleigh number Ra = 10(9). The cubical cavity of 0.7 m side-length is set to have a temperature difference of 39 K between the two facing cold and hot vertical walls. Experimentally measured temperature profiles for both the top and bottom walls were imposed as Dirichlet boundary conditions to implicitly account for wall-to-wall radiation effects. The focus of this research is the response of the subgrid-scale models and their ability to predict the flow in the thermally driven 3D cavity when the mesh resolution is coarse and below optimal LES standards. The research is motivated by investigating a feasible modeling strategy for particulate flows in the differentially heated cavity. As URANS and hybrid RANS-LES models fail to reproduce the flow accurately due to difficulty to model subtle physical mechanisms such as laminarization and three-dimensional effects, the alternative is LES applied on a coarse mesh. In a quantitative manner, the first and second-moment statistics are compared at different locations across the cavity against both previous fine mesh LES and experimental databases. For the first moment statistics, both models globally predict the flow fields well. However, for higher moments, the dynamic model outperforms the standard Smagorinsky model, and its predictions are globally in very good agreement with both reference LES and experimental measurements at a fraction of the CPU power needed for optimal LES. The presented results will motivate a further investigation of particle dispersion in the cubical cavity at moderate computational power.