AMBER Archive (2007)

Subject: RE: AMBER: Implicit solvent simulation is too slow.

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Fri Aug 10 2007 - 10:56:32 CDT

Hi Seung,
 
It tooks about 28.3 hours for 1ns simulation under following condition:

   28.3 hours / 1ns simulation
   16 Xeon DP 2.8 GHz CPUs
 
What is the interconnect here? If it is gigabit ethernet then you are
probably running slower than you would on a single node. You should
benchmark here to check. If it is infiniband or myrinet then it is probably
okay you should scale well with GB, especially if you use PMEMD.
 
   4107 atoms (288 residues protein + 1 ligand)
 
This is a big system for implicit solvent.
 
      5 ntpr = 100, ntwx = 100, ntwr = 1000,
 
Do you really need a restart file every 1000 steps? I would set this closer
to 10,000.
 
      7 ntb = 0,
      8 taup = 2.0,
 
taup does nothing when ntb=0.
 
      9 cut = 10,
 
Note this is a very small cutoff for an implicit solvent simulation. For
explicit solvent you can go as low as 8 angstroms due to PME but this is not
the case with GB. something like 16 or so would be better here - or
preferably no cut off at all. This will of course make your simulations even
slower.
 
I didn't run explicit solvent simulation yet. (It is waiting its turn in the
queue...)
But the same simulation done with GROMACS tell me it tooks about 15.4 hour s
for 1ns simulation.
 
I suspect you are comparing an explicit solvent simulation with an implicit
solvent one which is not a valid comparisson.
 
All the best
Ross
 
/\
\/
|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk <http://www.rosswalker.co.uk/> | PGP Key
available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

 

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