This contribution presents a parallel computational study of concrete fatigue using two formulations within the same potential-based constitutive framework: a mesoscale lattice discrete particle model (LDPM) explicitly representing heterogeneous microstructure through aggregate particles and cohesive interfaces, and a macroscale finite element microplane model (FEM-MP) employing a continuum directional material description. Both models derive fatigue laws from thermodynamic potentials [1], with damage driven by cumulative inelastic deformations—interface opening and sliding in LDPM, directional microplane strains in FEM-MP. Despite different spatial discretizations, the shared thermodynamic basis enables systematic comparison of their ability to reproduce structural-scale fatigue behavior. FEM-MP has been applied successfully to model fatigue-induced stress redistribution in prestressed concrete beams [2]. Model parameters are calibrated using standard uniaxial compression tests on concrete cylinders and then applied without modification to predict full-scale prestressed beam behavior under constant-amplitude compressive loading [3]. Digital image correlation (DIC) tracks compression zone damage evolution throughout the fatigue life, providing high-resolution experimental validation. Both models capture key experimental observations: three-stage fatigue creep, progressive damage localization, and stress redistribution due to microcracking and crushing through the structural depth. Comparison of material-scale (cylinder) and structural-scale (beam) fatigue characteristics—S-N curves and Sparks-Menzies relation—shows that structural members exhibit extended fatigue life due to stress redistribution, while the Sparks-Menzies slope remains essentially unchanged, confirming it as an intrinsic material property. Energy analysis indicates that total dissipated energy up to failure scales with the fatigue process zone rather than overall structural dimensions, representing a characteristic damage dissipation capacity per unit volume. The close agreement between LDPM and FEM-MP predictions, despite their distinct discretization approaches, validates the potential-based framework and supports the development of physics-based lifetime prediction methods beyond purely empirical relationships.