When only equilibrium thermodynamics or averages of position-dependent functions in classical molecular dynamics calculations are sought, the choice of the atomic masses in the Hamiltonian becomes irrelevant. Consequently, the masses can be used as free parameters to help improve conformational sampling efficiency in the spirit of Bennett's mass-tensor dynamics (Bennett, C. H. J. Comput. Phys. 1975, 19, 267.). Here we report on the effect of reducing side-chain and solvent masses on the folding behavior of a 9-residue hairpin-forming peptide in solution. A physical motivation for reducing the solvent mass is an effective reduction in solvent viscosity. Side-chain mass scaling is motivated by the idea of solvent potentials of mean force which are employed in select coarse-grained protein models. In the limit of very large mass differences, the mass reduction effectively creates an adiabatic decoupling between solvent, side-chain, and backbone motions, so that both the backbone and side chains move on the instantaneous solvent potential of mean force (PMF) surface, and the backbone additionally moves on the side-chain PMF. Because of the arbitrariness in the choice of masses, this limit only needs to be reached approximately in practice. In particular, we show that a 10-fold reduction in solvent masses and a side-chain mass scale that is intermediate between the scaled solvent and the backbone lead to a quantitative enhancement in conformational sampling.