From smallest to largest: (top) a single residue pair is
the smallest multi-body minimization possible, (middle) a full
protein conformation with a single flexible residue represents a small design,
(bottom) a full protein conformation with 20 flexible residues
represents a large design. For CPU hardware, concurrent minimizations correspond
to CPU threads. For GPU hardware, concurrent minimizations correspond to
streams defined by the CUDA framework. Faster minimization
speeds correspond with faster osprey runtimes. All minimizations were
performed on the Atx1 metallochaperone protein (PDB ID: 1CC8)47. Flexible residues were
modeled with continuous sidechain flexibility, and all other residues remained
completely fixed.