Namd on BGQ

Taking into consideration the official (test)[NAMD performance tuning recommendations], a parameter study was undertaken to test simulation performance and efficiency on the BG/Q cluster.

The following study was undertaken for a 246,000 atom membrane protein (Cytochrome c Oxdaise embedded in a TIP3P solvated DPPC bilayer) using the CHARMM36 forcefield (protein and lipids). Efficiency was measured with respect to the 16 ranks-per-node 512 core simulation.

Ranks	Cores	NAMD Config Options	ns/day	Efficiency
16	512		2.7874	1.00
16	1024		5.0483	0.91
16	1024	twoAwayX (default)	5.6204	1.01
16	2048	twoAwayX (default)	10.0699	0.90
16	2048	twoAwayXY	10.5875	0.95
16	4096	twoAwayX	14.3285	0.64
16	4096	twoAwayXY (default)	17.6313	0.79
16	4096	twoAwayXYZ	16.7962	0.75
16	8192	twoAwayX	23.5198	0.53
16	8192	twoAwayXY (default)	24.9981	0.56
16	16384	twoAwayX	23.6706	0.27
16	16384	twoAwayXY	28.3096	0.32
16	16384	twoAwayXYZ (default)	27.9791	0.31
16	4096	twoAwayXY, Pencils8, lblUnload=yes	12.9322	0.58
16	4096	twoAwayXY, Pencils12, lblUnload=yes	17.2722	0.77
16	4096	twoAwayXY, Pencils16, lblUnload=yes	16.0208	0.72
16	4096	twoAwayXY, Pencils20, lblUnload=yes	15.4154	0.69
16	4096	twoAwayXY, Pencils12	16.2079	0.73
16	4096	twoAwayXY, Pencils16	17.9231	0.80
16	4096	twoAwayXY, Pencils20	17.9949	0.81
16	4096	twoAwayXY, Pencils24	17.8308	0.80
16	4096	twoAwayXY, Pencils36	16.9666	0.76
8	4096	twoAwayXY, Pencils20	18.2473	0.82
16	4096	twoAwayXY, Pencils20	17.9949	0.81
32	4096	twoAwayXY, Pencils20	13.9414	0.63
4	512		2.8589	1.03
8	512		2.8367	1.02
16	512		2.7874	1.00
32	512		2.2951	0.82
16	512	unloadZero	2.7932	1.00

Documentation

1. NAMD 2.9 User Guide
2. NAMD Performance Tuning Wiki