In order to keep people from using implicit solvent to sample protein conformations, I highly encourage to use at least one of the latest REMD adaptions like TIGER2A which allow sampling in explicit solvent with reduces computational resources.
Even less resources are required for hybrid solvent approaches like TIGER2h or TIGER2hs with much better accuracy than pure implicit solvent simulation.
Article TIGER2 with solvent energy averaging (TIGER2A): An accelerat...
Article Replica Based Protein Structure Sampling Methods: Compromisi...
Article Replica-Based Protein Structure Sampling Methods II: Advance...
From our experience (to be published) also hamiltonian based approaches like REST2 or dihedral gREST seem not to be any alternative, since they sample a rather different universe (see image attached)