Model-based predictive control (MPC) refers to a class of advanced control strategies, which has been widely used in the process industry through distributed system architectures. With advances in hardware optimization and development techniques, MPC implementations in embedded systems have become intensive, however, important properties, such as ensuring stability and feasibility, still represent an open problem for real-time applications. . This work presents the implementation in embedded systems of a predictive controller with guaranteed nominal stability and feasibility, whose formulation results in a quadratic programming (QP) problem. The efficient ADMM (Alternating Direction Method of Multipliers) optimization technique was used to solve the QP, providing a sufficiently accurate solution in a few iterations. For that, o ine simulations were performed to properly tune the parameters of the controller and ADMM. The control of a fast dynamic electromechanical system was implemented in a CompactLogix L32E programmable logic controller (PLC), whose limited memory and processing resources impose challenges for the application, and in an ESP32 microcontroller, which has high processing power and memory. The results include hardware-in-the-loop simulations and experimental tests at the plant. The implementation in the PLC demanded about 40% of the total available memory and the computation time of the problem was approximately 90 milliseconds. On the other hand, in the microcontroller, only 18% of the total available memory was used, while the computation time of the problem was approximately 5 milliseconds. The results show the effectiveness of the controller in both hardware for the case of reference tracking, even in the face of model uncertainties. However, in large problems or where ADMM requires a greater number of iterations, the application may require more advanced features than those available on conventional PLCs.