We studied the sensitivity of the energetic and geometrical features of the proline ring (pyrrolidine) to the quantum mechanical computational approach by adopting the proline monomer, trimer, and polymer, as simplified collagen protein models. Within the Density Functional Theory (DFT) approach, we tested the ability of different functionals (GGA PBE and the hybrid B3LYP), added with a posteriori empirical dispersion corrections (D), to predict the conformational potential energy surface of the five-membered heterocycle pyrrolidine ring for the above models, dictating the collagen main features. We also compared the DFT-D results with those from the recently proposed cost-effective HF-3c method and our variant HF-3c-027, both based on Hartree-Fock Hamiltonian and Gaussian minimal basis set properly corrected for basis set superposition error, structure deficiencies, and dispersion interactions. We found that dispersion interactions are essential to destabilize specific conformers. While the HF-3c and its HF-3c-027 variant are unreliable to predict accurately the energy of the ring conformers, structures are accurate. Indeed, the cost-effective DFT-D//HF-3c-027 approach in which the energetic is from the accurate DFT-D method on HF-3c-027 structures provides energetic in line with that derived by the costly DFT-D//DFT-D approach, paving the way to simulate more realistic collagen models of much larger size. The adoption of either PBE or B3LYP functional for the electronic part of the DFT-D method gives very similar results, recommending the first as the most cost-effective method for simulating large collagen models. The predicted most stable conformation computed for the periodic poly proline (type II) model allows for a higher flexibility, in agreement with experimental studies on collagen protein. The present results open the way to large-scale calculations of the collagen/hydroxyapatite system, crucial for understanding the atomistic details in bones and teeth.