The paper critically discusses Large-Eddy Simulation (LES) potential to investigate cycle-to-cycle variability (CCV) in internal combustion engines. Particularly, the full load/peak power engine speed operation of a high-performance turbocharged GDI unit, for which ample cycle-to-cycle fluctuations were observed during experimental investigations at the engine test bed, is analyzed through a multi-cycle approach covering 25 subsequent engine cycles. In order to assess the applicability of LES within the research and development industrial practice, a modeling framework with a limited impact on the computational cost of the simulations is set up, with particular reference to the extent of the computational domain, the computational grid size, the choice of boundary conditions and numerical sub-models [1, 2, 3]. In order to evaluate the applicability of the adopted approach to the resolution of an adequate portion of the overall turbulent energy spectrum, different grid metrics are at first introduced, based on criteria available in literature [4, 5]. A qualitative comparison between CFD results and experimental evidence is then carried out in terms of both in-cylinder pressure envelope and coefficients of variation for any of indicated mean effective pressure, 10%, 50% and 90% of fuel burnt distributions among the investigated cycles. Particularly, a detailed analysis of the physical factors influencing the exhibited cycle-to-cycle variability is performed through the use of correlation coefficients, aiming at highlighting possible hierarchies between the many involved phenomena and the observed engine behavior. Finally, a phase-dependent Proper Orthogonal Decomposition (POD). Particularly, while POD applications available in literature mainly cover vector fields and flow structures [6, 7], in the present paper the analysis is extended to scalar fields describing the combustion process evolution and its cyclic variability, and results are critically analyzed and commented.