AMBER Archive (2006)

Subject: Re: AMBER: Different results between serial and parallel runs in AMBER 8

From: Carlos Simmerling (carlos.simmerling_at_stonybrook.edu)
Date: Mon Mar 06 2006 - 18:33:14 CST

try not using Langevin (set ntt=1).
have you applied all bug fixes?
it may be that the random number sequence isn't the same for
multiple cpus.

Ilyas Yildirim wrote:

>Dear All,
>
>I have a question on parallel and serial MD runs in AMBER 8. I have
>created a system, which is a 4X4 base paired RNA, solvated in water. I
>have created ordinary prmtop and inpcrd files. I compiled AMBER 8 both in
>serial and parallel versions (renamed them as sander.serial and
>sander.parallel). Then, I did a short run on this system, using
>sander.serial and sander.parallel. Here are the first two step results:
>
>--------------------------- parallel run -------------------------------
>.
>.
>.
> 10 fs simulation
> &cntrl
> imin=0,
> ntx=1,irest=0,
> ntpr=1,ntwr=1,ntwx=1,
> ntc=2,ntf=2,ntb=1,cut=8,
> igb=0,
> ntr=0,
> nstlim=10,dt=0.001,nscm=5000,nrespa=1,
> ntt=3,gamma_ln=1,tempi=300,temp0=300,ig=233,
> ntp=0,taup=2.0,pres0=1
> /
>.
>.
>.
> NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 439.42 PRESS = 0.0
> Etot = 4337.6493 EKtot = 6803.5776 EPtot = -2465.9282
> BOND = 66.7991 ANGLE = 128.2466 DIHED = 205.4474
> 1-4 NB = 74.6955 1-4 EEL = -1233.7386 VDWAALS = 100.1044
> EELEC = -1807.4827 EHBOND = 0.0000 RESTRAINT = 0.0000
> Ewald error estimate: 0.1351E-02
> ------------------------------------------------------------------------
> NSTEP = 1 TIME(PS) = 0.001 TEMP(K) = 345.43 PRESS = 0.0
> Etot = 2882.4294 EKtot = 5348.3577 EPtot = -2465.9282
> BOND = 66.7991 ANGLE = 128.2466 DIHED = 205.4474
> 1-4 NB = 74.6955 1-4 EEL = -1233.7386 VDWAALS = 100.1044
> EELEC = -1807.4827 EHBOND = 0.0000 RESTRAINT = 0.0000
> Ewald error estimate: 0.1351E-02
>.
>.
>.
>-------------------------------------------------------------------------
>
>------------------------------ serial run -------------------------------
>.
>.
>.
> 10 fs simulation
> &cntrl
> imin=0,
> ntx=1,irest=0,
> ntpr=1,ntwr=1,ntwx=1,
> ntc=2,ntf=2,ntb=1,cut=8,
> igb=0,
> ntr=0,
> nstlim=10,dt=0.001,nscm=5000,nrespa=1,
> ntt=3,gamma_ln=1,tempi=300,temp0=300,ig=233,
> ntp=0,taup=2.0,pres0=1
> /
>.
>.
>.
> NSTEP = 0 TIME(PS) = 0.000 TEMP(K) = 439.42 PRESS = 0.0
> Etot = 4337.6493 EKtot = 6803.5776 EPtot = -2465.9282
> BOND = 66.7991 ANGLE = 128.2466 DIHED = 205.4474
> 1-4 NB = 74.6955 1-4 EEL = -1233.7386 VDWAALS = 100.1044
> EELEC = -1807.4827 EHBOND = 0.0000 RESTRAINT = 0.0000
> Ewald error estimate: 0.1351E-02
>-------------------------------------------------------------------------
> NSTEP = 1 TIME(PS) = 0.001 TEMP(K) = 332.09 PRESS = 0.0
> Etot = 2675.8869 EKtot = 5141.8152 EPtot = -2465.9282
> BOND = 66.7991 ANGLE = 128.2466 DIHED = 205.4474
> 1-4 NB = 74.6955 1-4 EEL = -1233.7386 VDWAALS = 100.1044
> EELEC = -1807.4827 EHBOND = 0.0000 RESTRAINT = 0.0000
> Ewald error estimate: 0.1351E-02
>.
>.
>.
>-------------------------------------------------------------------------
>
>I have used the same md.in file. In the above results, the NSTEP = 0 have
>the same values (which is expected because the structures are exactly the
>same). But when it becomes NSTEP = 1, the temp. values changes; in one
>case it is 345.43 and in the other case it is 332.09. And as a result, the
>rest of the NSTEPs have different values when serial and parallel runs are
>compared. Is this an expected behavior?
>
>Best,
>
>
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu