[gmx-users] parallel mdrun (mpimdrun) is slower than non-parallel mdrun?

[wuxiao]

Dear gmx users,

  I am performing a MD simulation of a polymer film using GROMACS-4.0.7. My computer has a four-core CPU, which enables the parallel implementation. At first, I used mpimdrun to simulate the system, only to find with -v that the vol is too small (about 0.2) and the imb F is too big (over 150%) so about 4 days is required for finishing 4ns MD. However, when I turn to non-parallel mdrun, only half of the time (about 2 days) is required for finishing the same task. Is it reasonable? How to explain it?

  Thanks a lot for any replies to this post

 

  Chaofu Wu, Dr.
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20100730/89507874/attachment.html>