Skip to main content
. 2019 Jul 3;117(3):602–612. doi: 10.1016/j.bpj.2019.06.030

Table 2.

Energy Evaluation Timings for Global Side-Chain Optimization of ACTG1 Residues 6–375 and COCH Residues 27–125 Using a Varying Number of GPUs

Number of Nodes Number of GPUs Time for Energies (sec)
Speed-Up (Relative to Using All CPU Cores)
ACTG1 COCH ACTG1 COCH
1 0 (CPUs only) 33,126 5505 1.0× 1.0×
1 1 2576 479 12.9× 11.5×
1 2 1277 251 25.9× 21.9×
1 4 656 142 50.5× 38.8×
2 8 336 76 98.6× 72.4×
4 16 175 43 189.3× 128.0×

Each node contains two Intel Xeon E5-2680v4 CPUs and four NVIDIA GTX 1080 TI GPUs.