<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
  </head>
  <body text="#000000" bgcolor="#ffffff">
    In addition, you're only updating your neighbourlist every 40 ps. If
    you're going to use a 4 fs timestep, I suggest that you use
    nstlist=5. Also, you appear to not be using any constraints while
    you are using a 4 fs timestep. <br>
    <br>
    I suggest that you stop worrying about why the run is slower and do
    some tutorials and read the manual and ensure that you understand
    the consequences of your .mdp options.<br>
    <br>
    Chris.<br>
    <br>
    -- original message --<br>
    <pre>Denny Frost wrote:
&gt;<i> about 12000 atoms, 8 nodes, CentOS 5.3/Linux, infiniband.  Below is a 
</i>&gt;<i> copy of my mdp file.
</i>&gt;<i> 
</i>&gt;<i> title               =  BMIM+PF6
</i>&gt;<i> cpp                 =  /lib/cpp
</i>&gt;<i> constraints         =  all_bonds
</i>&gt;<i> integrator          =  md
</i>&gt;<i> dt                  =  0.004   ; ps !
</i>&gt;<i> nsteps              =  20000000   ; total 4ns.
</i>&gt;<i> nstcomm             =  1
</i>&gt;<i> nstxout             =  50000
</i>&gt;<i> nstvout             =  50000
</i>&gt;<i> nstfout             =  0
</i>&gt;<i> nstlog              =  5000
</i>&gt;<i> nstenergy           =  5000
</i>&gt;<i> nstxtcout           =  25000
</i>&gt;<i> nstlist             =  10
</i>&gt;<i> ns_type             =  grid
</i>&gt;<i> pbc                 =  xyz
</i>&gt;<i> coulombtype         =  PME
</i>&gt;<i> vdwtype             =  Shift
</i>&gt;<i> rlist               =  1.0
</i>&gt;<i> rcoulomb            =  1.0
</i>&gt;<i> rvdw                =  1.0
</i>&gt;<i> fourierspacing      =  0.6
</i>
This fourierspacing is 5-6 times larger than what is normally accepted as 
sufficiently accurate.  A sparse grid will make the PME algorithm faster, 
actually, but at the expense of accuracy.

Can you post the domain decomposition statistics from the .log file?  They 
appear just above the energies from time 0.  What did grompp tell you about the 
relative PME:PP load?

-Justin

&gt;<i> ;pme_order           =  4
</i>&gt;<i> ewald_rtol          =  1e-5
</i>&gt;<i> ; Berendsen temperature coupling is on in two groups
</i>&gt;<i> Tcoupl              =  berendsen
</i>&gt;<i> tc_grps             =  BMI      PF6      
</i>&gt;<i> tau_t               =  0.1  0.1
</i>&gt;<i> ref_t               =  300  300
</i>&gt;<i> nsttcouple          =  1
</i>&gt;<i> ; Energy monitoring
</i>&gt;<i> energygrps          =  BMI      PF6
</i>&gt;<i> ; Isotropic pressure coupling is now on
</i>&gt;<i> Pcoupl              =  berendsen
</i>&gt;<i> pcoupltype          =  isotropic
</i>&gt;<i> ;pc-grps             =  BMI      PFF
</i>&gt;<i> tau_p               =  1.0
</i>&gt;<i> ref_p               =  1.0
</i>&gt;<i> compressibility     =  4.5e-5
</i>&gt;<i> 
</i>&gt;<i> ; Generate velocites is off at 300 K.
</i>&gt;<i> gen_vel             =  yes
</i>&gt;<i> gen_temp            =  300.0
</i>&gt;<i> gen_seed            =  100000
</i>&gt;<i> 
</i>&gt;<i> 
</i>&gt;<i> On Thu, Jan 27, 2011 at 4:12 PM, Dallas Warren &lt;<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">Dallas.Warren at monash.edu</a> 
</i>&gt;<i> &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">Dallas.Warren at monash.edu</a>&gt;&gt; wrote:
</i>&gt;<i> 
</i>&gt;<i>     You will need to provide more details on the system.  How many
</i>&gt;<i>     atoms, what sort of computer system is it being run on, how many
</i>&gt;<i>     nodes, copy of the mdp file etc.
</i>&gt;<i> 
</i>&gt;<i>      
</i>&gt;<i> 
</i>&gt;<i>     Catch ya,
</i>&gt;<i> 
</i>&gt;<i>     Dr. Dallas Warren
</i>&gt;<i> 
</i>&gt;<i>     Medicinal Chemistry and Drug Action
</i>&gt;<i> 
</i>&gt;<i>     Monash Institute of Pharmaceutical Sciences, Monash University
</i>&gt;<i>     381 Royal Parade, Parkville VIC 3010
</i>&gt;<i>     <a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">dallas.warren at monash.edu</a> &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">dallas.warren at monash.edu</a>&gt;
</i>&gt;<i> 
</i>&gt;<i>     +61 3 9903 9304
</i>&gt;<i>     ---------------------------------
</i>&gt;<i>     When the only tool you own is a hammer, every problem begins to
</i>&gt;<i>     resemble a nail.
</i>&gt;<i> 
</i>&gt;<i>      
</i>&gt;<i> 
</i>&gt;<i>     *From:* <a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-bounces at gromacs.org</a>
</i>&gt;<i>     &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-bounces at gromacs.org</a>&gt;
</i>&gt;<i>     [mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-bounces at gromacs.org</a>
</i>&gt;<i>     &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-bounces at gromacs.org</a>&gt;] *On Behalf Of *Denny Frost
</i>&gt;<i>     *Sent:* Friday, 28 January 2011 9:34 AM
</i>&gt;<i>     *To:* Discussion list for GROMACS users
</i>&gt;<i>     *Subject:* [gmx-users] Slow Runs
</i>&gt;<i> 
</i>&gt;<i>      
</i>&gt;<i> 
</i>&gt;<i>     I am taking over a project for a graduate student who did MD using
</i>&gt;<i>     Gromacs 3.3.3.  I now run similar simulations with Gromacs 4.5.1 and
</i>&gt;<i>     find that they run only about 1/2 to 1/3 as fast as the previous
</i>&gt;<i>     runs done in Gromacs 3.3.3.  The runs have about the same number of
</i>&gt;<i>     atoms and both use opls force fields.  The mdp files is virtually
</i>&gt;<i>     the same (I copied them).  The only major difference is that my runs
</i>&gt;<i>     have difference species and thus have different (although smaller)
</i>&gt;<i>     itp files.  The runs are stable and give reasonable thermodynamic
</i>&gt;<i>     properties - they're just slow.  Has anyone had any experience with
</i>&gt;<i>     something like this?
</i>&gt;<i> 
</i>&gt;<i> 
</i>&gt;<i>     --
</i>&gt;<i>     gmx-users mailing list    <a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users at gromacs.org</a>
</i>&gt;<i>     &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users at gromacs.org</a>&gt;
</i>&gt;<i>     <a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">http://lists.gromacs.org/mailman/listinfo/gmx-users</a>
</i>&gt;<i>     Please search the archive at
</i>&gt;<i>     <a href="http://www.gromacs.org/Support/Mailing_Lists/Search">http://www.gromacs.org/Support/Mailing_Lists/Search</a> before posting!
</i>&gt;<i>     Please don't post (un)subscribe requests to the list. Use the
</i>&gt;<i>     www interface or send it to <a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-request at gromacs.org</a>
</i>&gt;<i>     &lt;mailto:<a href="http://lists.gromacs.org/mailman/listinfo/gmx-users">gmx-users-request at gromacs.org</a>&gt;.
</i>&gt;<i>     Can't post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists">http://www.gromacs.org/Support/Mailing_Lists</a>
</i>&gt;<i> 
</i>&gt;<i> 
</i>
-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
<a href="http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin">http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin</a>
</pre>
  </body>
</html>