The phenomenon of variation of fluid properties with depth has been observed in reservoirs with large thickness or high dip angle where primarily the gravitational forces along with the thermal effect play an important role on the overall composition of the system. Compositional grading is more evident in high shrinkage oils, volatile oils, rich gas condensates as well as the black oils. Compositional grading is a key issue in dynamic modeling of the reservoir and must be given enough care during both estimation of the stock-tank initially in place and reservoir performance. This paper presents an engineering evaluation of fluid characterization of a heavy, viscous oil from an Iranian oil reservoir which is considered for a gas injection project. After screening of all available PVT data, the validated data were input into the PVT package and reproduced all existing experiments with only one EOS. Since there was no firm evidence of different equilibrium regions inside the reservoir, it was decided to have one EOS for the whole reservoir with varying mole fraction with increasing depth. Viscosity is the main property to be predicted with EOS in this task. Mole fractions of C7+ are critical in this case and need special attention. For regressing experimental viscosity, two regions were defined. The results show that the suggested EOS will lead to a reasonable match of all experimental data and when was used in the reservoir simulation in the next phase of the project, the initial oil in place was close to that obtained based on previous black oil model.