Re: [AMBER] Intel Xeon Phi 7120P

From: Elvis Martis <elvis.martis.bcp.edu.in>
Date: Wed, 15 Mar 2017 07:03:44 +0000

HI,
One thing you can do is double the time step to 4 fs (dt=0.004) using the hydrogen mass repartitioning (you can do it using parmed).
And if it is not necessary you can perform you MD using NVT ensemble which is faster than NPT (you are currently doing NPT).

    Best Regards



Elvis Martis
Ph.D. Student (Computational Chemistry)
 at Bombay College of Pharmacy

A  Kalina, Santacruz [E], Mumbai 400098, INDIA
W www.elvismartis.in
Skype. adrian_elvis12




-----Original Message-----
From: Sofia Vasilakaki [mailto:svasilak.chem.uoa.gr]
Sent: Tuesday, March 14, 2017 7:57 PM
To: amber.ambermd.org
Subject: [AMBER] Intel Xeon Phi 7120P

Hi everyone,

I am attempting to run MD simulation of a 262,572 atoms system (protein attached to a lipid membrane) on a phi 7120P.

It takes 38h for 2ns and I am wondering how I could accelerate things.

I use:
pmemd.mic_offload.MPI
nodes=1
ntasks-per-node=20
cpus-per-task=1
gres=mic:1

and the input file is:
   imin=0, ntx=5, irest=1,
   ntc=2, ntf=2, tol=0.0000001,
   nstlim=1000000, ntt=3, gamma_ln=1.0,
   temp0=310.0,
   ntpr=20000, ntwr=500000, ntwx=20000,
   dt=0.002, ig=-1,
   ntb=2, ntp=2, cut=10.0, barostat=1, ioutfm=1, ntxo=2

Please, let me know if you need more info about my system.

Thank you in advance.

Cheers,
Sofia



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Mar 15 2017 - 00:30:03 PDT
Custom Search