The paper presents an alternative robust model predictive control strategy based on simultaneous exploitation of two principles: optimization of the nominal system convergence, and additional saturation of control inputs using weights on saturated and non-saturated values. The strategy decreases overall computational complexity and ensures control performance comparable with the original strategy. Moreover, feasibility set of initial conditions is larger. The properties of proposed method were investigated using simulation case-study of uncertain and open-loop unstable chemical reactor stabilization.