Three-dimensional reactive computational fluid dynamics (CFD) plays a crucial role in IC engine development tasks complementing experimental efforts by providing improved understanding of the combustion process. A widely adopted combustion model in the engine community for (partially) premixed combustion is the G-Equation where the flame front is represented by an iso-level of an arbitrary scalar G. A convective-reactive equation for this iso-surface is solved, for which the turbulent flame speed ST must be provided. In this study, the commonly used and well-established Damköhler approach is compared to a novel correlation, derived from an algebraic closure for the scalar dissipation of reaction progress as proposed by Kolla et al. . The predictions from the two correlations are probed towards their sensitivities by means of experimental data from two distinctly different engine configurations: 1) a lean burn spark ignition natural gas engine for power generation, derived from a flat head/bowl-in piston compression ignition engine with roughly two liters of displacement per cylinder, and, 2) a small bore 250 cc single cylinder port fuel injected gasoline engine with a typical four-valve pent roof arrangement. The sensitivity of the flame speed closures towards the in-cylinder turbulent flow field is investigated for a sweep in turning speed for both engines close to full load. The numerical predictions in terms of pressure trace as well as heat release rates for both engines are compared to experimental test bench data and the predictive capabilities of the proposed closures for the two considerably different engine scales are studied. Additionally, the impact of two different RANS turbulence models (k-ε RNG and standard k-ω) on the results is discussed. The combustion regime is further classified in the Borghi diagram throughout the combustion event, where the flame speed closures are further investigated with emphasis on the early and late combustion stages.