Fracture of polymer-metal interfaces is one of the main failure modes occurring in micro-electronic components. This phenomenon is particularly true when considering the delamination of several layers of an interconnect structure. In order to predict the failure nucleation and the crack propagation into the composite material, the finite element analysis is one of the key procedures. Even though simple linear models have been considered for years, we are now facing the necessity of using more complex models including non-linearity which can occur, in this case, with the presence of high local stresses near the crack front. However, the computational time can sometimes be incredibly long. Moreover, the fact that the considered materials are quasi-brittle brings some numerical difficulties such as sharp snap-back and snap-through. The actual challenge resides in obtaining a reliable result in a reasonable time of calculation. The present work considers the implementation of a new non-linear finite element solver, developed for the MSc. Marc/Mentat package software. It is based on a general arc-length constraint which considers the energy released during the propagation of the crack. This offers the advantage of being directly linked to the failure process, and no previous knowledge of the failure behavior is required. The models considered in this work represent the simulation of crack propagations in multilayer electronic systems, such as SIP devices, and are based on a cohesive zone approach. In order to clearly understand the issues of this problem, this work includes a brief description of the fracture mechanics and reviews the existing nonlinear finite element solvers. After explaining the principle of the energy release solver and the different issues due to its implementation, its efficiency is compared to pre-implemented solvers, such as the Crisfield method. The results show a significant improved robustness of the new energy released method compared to the previous arc-length methods.