On the time-scale of a density functional calculation, these empirical dispersion corrections require only a minuscule time, yet significantly improve the accuracy of the relative energies. Thus, even though this work is concerned with intramolecular interactions in conformers, dispersion-corrected density functional calculations should always be used. Continued efforts, such as the improved D3 methods\cite{Witte_2017} or the new D4 method\cite{Caldeweyher_2019,Caldeweyher_2017} will hopefully improve their accuracy further.

Comparison of Timing

As discussed above, a frequent concern for conformer screening is the relative computational performance. In general, classical molecular force field methods have been preferred since they allow the generation of hundreds of conformers per compound in seconds. While traditional high-level ab initio methods are considered a "gold standard" for thermochemical energies, the time required for a single point energy evaluation may be high. For this work, all timings are single-core CPU times using a 2.60 GHz Intel Skylake CPU (Intel Xeon Gold 6126) with 192GB RAM per node.
As indicated in Figure \ref{413433}, hybrid density functional methods such as B3LYP-D3BJ require significant single-computational time for single-point energies of medium-sized organic molecules (median 26±0.3 minutes) compared to GGA methods such as PBE or approximate density functional tight binding methods such as GFN1 / GFN2 (median 2.6±0.06 s yields ~600x speedup). Conventional density functional methods nevertheless represent a meaningful mid-point relative to DLPNO-CCSD(T) method, which may be faster than traditional coupled cluster methods but are still five to ten times slower than B3LYP (i.e., hours per single point energy).