Skip to main content
. 2019 May 3;33(6):531–558. doi: 10.1007/s10822-019-00203-1

Table 3.

Summary of ForceGen calculation speed using different hardware and core/thread combinations, with throughput calculated assuming molecule-level parallelism in the 36-core 1-thread scenario (36 simultaneous single-threaded jobs)

Method N cores N threads Median time (s) Mean time (s) N jobs Molecules per hour Molecules per day
FGen -pfastf 36 36 0.30 0.62 1 5806
FGen -pfastf 4 8 0.52 0.93 1 3865
FGen -pfastf 36 1 0.59 1.35 36 96,000 2,304,000
FGen -pscreen 36 36 0.48 1.07 1 3364
FGen -pscreen 4 8 0.92 1.69 1 2133
FGen -pscreen 36 1 1.28 3.03 36 42,772 1,026,534
FGen -pgeomf 36 36 0.61 1.53 1 2352
FGen -pgeomf 4 8 1.24 2.61 1 1382
FGen -pgeomf 36 1 1.62 4.45 36 29,123 698,966
FGen -pgeom 36 36 0.88 2.89 1 1246
FGen -pgeom 4 8 2.00 6.32 1 570
FGen -pgeom 36 1 3.01 15.11 36 8577 205,850