We present an evaluation of our newly developed density functional theory (DFT)-D3 dispersion correction D3(CSO) in comparison to its predecessor D3(BJ) for geometry optimizations. Therefore, various benchmark sets covering bond lengths, rotational constants, and center of mass distances of supramolecular complexes have been chosen. Overall both corrections give accurate structures and show no systematic differences. Additionally, we present an optimized algorithm for the computation of the DFT-D3 gradient, which reduces the formal scaling of the gradient calculation from O (N-3) to O (N-2). Published by AIP Publishing.