Efficient exploration of the conformational space of peptides embedded in biological membranes is vital to extract converged thermodynamic and kinetic data from simulation studies. Recently developed implicit membrane models promise vast increases in sampling efficiency compared to explicit membrane simulations, allowing for ab initio structure prediction and functional studies. In this study, a previously developed implicit membrane model, based on the generalized Born method, is compared to an explicit di-palmitoyl-phosphatidyl-choline lipid bilayer and an octane slab membrane mimic. The complete folding process of a synthetic 16-residue peptide is compared using these three setups. Since the comparison requires the entire folding pathway to be captured, individual simulations ranged up to 3 mus of MD. A quantitative sampling comparison using a wide range of performance metrics reveals that the implicit membrane model is at least 2 orders of magnitude more efficient than the simplest explicit setups.