The structure of a protein is often not completely accessible by experiments. In silico, replica exchange molecular dynamics (REMD) is the standard sampling method for predicting the secondary and tertiary structures from the amino acid sequence, but it is computationally very expensive. Two recent adaptations from REMD, temperature intervals with global exchange of replicas (TIGER2) and TIGER2A, have been tested here in implicit and explicit solvents. Additionally, explicit, implicit, and hybrid solvent REMD are compared. On the basis of the hybrid REMD (REMDh) method, we present a new hybrid TIGER2h algorithm for faster structural sampling, while retaining good accuracy. The implementations of REMDh, TIGER2, TIGER2A, and TIGER2h are provided for nanoscale molecular dynamics (NAMD). All the methods were tested with two model peptides of known structure, (AAQAA)3 and HP7, with helix and sheet motifs, respectively. The TIGER2 methods and REMDh were also applied to the unknown structure of the collagen type I telopeptides, which represent bigger proteins with some degree of disorder. We present simulations covering more than 180 μs and analyze the performance and convergence of the distributions of states between the particular methods by dihedral principal component and secondary structure analysis.