Quantifying the cell permeability of cyclic peptides is crucial for their rational drug design. However, the reasons remain unclear why a minor chemical modification, such as the difference between Ras inhibitors cyclorasin 9A5 and 9A54, can substantially change a peptide's permeability. To address this question, we performed enhanced sampling simulations of these two 11-mer peptides using the coupled Nosé-Hoover equation (cNH) we recently developed. The present cNH simulations realized temperature fluctuations over a wide range (240-600 K) in a dynamic manner, allowing structural samplings that were well validated by nuclear Overhauser effect measurements. The derived structural ensembles were comprehensively analyzed by all-atom structural clustering, mapping the derived clusters onto principal components (PCs) that characterize the cyclic structure, and calculating cluster-dependent geometric and chemical properties. The planar-open conformation was dominant in aqueous solvent, owing to inclusion of the Trp side chain in the main-chain ring, while the compact-closed conformation, which favors cell permeation due to its compactness and high polarity, was also accessible. Conformation-dependent cell permeability was observed in one of the derived PCs, demonstrating that decreased cell permeability in 9A54 is due to the high free energy barrier separating the two conformations. The origin of the change in free energy surface was determined to be loss of flexibility in the modified residues 2-3, resulting from the increased bulkiness of their side chains. The derived molecular mechanism of cell permeability highlights the significance of complete structural dynamics surveys for accelerating drug development with cyclic peptides.