From: Jacqueline Schmidt (jacqueline_schmidt_1983_at_yahoo.com)
Date: Tue Dec 25 2012 - 17:29:05 CST
Dear Namd users,
I have cam across a weird discrepancy in namd results on different machines:
I have run a test system on tree clusters and I got different results as below
1-red hat 64 bit (used 12 cpus)
ETITLE: TS BOND ANGLE DIHED IMPRP ELECT VDW BOUNDARY MISC KINETIC TOTAL TEMP TOTAL2 TOTAL3 TEMPAVG PRESSURE GPRESSURE VOLUME PRESSAVG GPRESSAVG
ENERGY: 2000 6281.3116 3470.8004 36.9492 2.4299 -205400.7859 1323.2906 0.0000 0.0000 5776.4835 -188509.5208 273.2524 -188466.2680 -188470.2608 267.9630 1183.8851 490.4254 78989.2992 36.6953 31.7657
2-red hat 64 bit (used 12 cpus) - another cluster than above
ENERGY: 2000 6231.8194 3307.1326 40.0257 2.9600 -205430.4443 1463.4158 0.0000 0.0000 5632.5355 -188752.5553 266.4430 -188710.4544 -188711.2469 264.3466 168.5647 69.4911 78739.9095 -63.1504 -57.8641
3-red hat 64 bit (machine is accessorized with gpu, here I used the same executable file as I did on the other machines for initial test)
ENERGY: 2000 6238.3632 3279.5853 38.7594 6.3686 -205074.7611 1388.3075 0.0000 0.0000 5733.3590 -188390.0181 271.2124 -188345.6856 -188346.5149 271.4772 -409.1661 -111.6605 78769.2246 3.6113 3.6517
The differences are very large (at 2 ps simulation); specially for pressure(the differences are really dramatic)
Is this something than others have observed? I have been using the available executable file, shouldn't I? I assumed it should work fine!!!
I would appreciate any comment on this please
Thanks and merry Christmas!
jackie
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:52 CST