Steady and unsteady natural convection heat transfer in a nanofluid-filled square enclosure cavity with various heat source configurations at the bottom were numerically studied using a characteristic-based split scheme finite element method. A wide range of Rayleigh numbers , solid volume fractions of nanoparticles , the lengths , and locations of heat sources as well as different types of nanoparticles were considered. Results showed that addition of nanoparticles into the base fluid leads to evident enhancement of heat transfer, particularly at low Rayleigh numbers and that cooling performance is strongly influenced by thermo-physical properties of the nanoparticles. Temporal development of thermal layer in a nanofluid-filled cavity was demonstrated from initial to steady stage through unsteady analysis. We found that the maximum temperature and the average Nusselt number started to oscillate, which is a typical characteristics of high Rayleigh number flow, at higher Rayleigh number in nanofluid than in pure water.