Modern industrial plants consist of many processes occurring simultaneously which are to be monitored and controlled at all time, during plant operation. Typically, industrial plants are multi-input multi-output type and they also exhibit strong non-linear dynamics. Three tank liquid level control is such a laboratory based benchmark multi-variable control problem for modeling, identification, fault detection & diagnosis and fault tolerant control system design. Multi-variable processes make monitoring and control difficult for conventional controllers. To overcome this problem most modern industries, use a model based controller as supervisory controller. However, these controllers are designed based on linear model of the process. Linear models often fail to capture the non-linear dynamics which results in the failure of control systems. In this article, we compared the performance of Linear Quadratic Gaussian Control (LQG) with the Non-linear Model Predictive Control (NMPC) to achieve the servo plus disturbance rejection and regulatory control of a three tank system in presence of changing valve position which serves as the disturbance input. As anticipated, the simulation results reveal better competency of NMPC over LQG in handling disturbances.