In this study, the airflow in a room that contains a heat source is simulated numerically. The flow is considered turbulent and buoyant. The results of the mathematical model are validated with available experimental data at specific locations in the domain. A simple geometry is adopted, consisting of a room with a door that plays the role of both inlet-outlet for the fluid (air). At the centre of the room a methane burner is placed to serve as a heat source. The problem is simulated using two turbulence models, the well-known standard k-epsilon model and the RNG k-epsilon. model, both modified to account for buoyancy effects on turbulence. The burner is considered as a volumetric heat source. It is concluded that the fire plume development as well as the distributions of velocity and temperature are reasonably well predicted. Following this conclusion, both models are also applied to a different, more complex geometry that consisted of two rooms communicating via a door, while the heat source was placed in the first room. Unfortunately, there are no experimental data to compare with for this case, but the results appear plausible. Finally, important design factors, such as mass flow rates and neutral-plane heights, are calculated utilizing the CFD results, and are compared with those obtained by well-known empirical correlations. It is concluded that the bi-directional flow existing through the burning-room vent is similarly predicted by both turbulence models; the RNG k-epsilon model leading to higher, and more accurate predictions of temperature variations within the hot upper layer, at least for the single-room case. (c) 2009 Elsevier Ltd. All rights reserved.