CP2K 4.1 versus CP2K 2.1

[jim wang]

Hi !

I am a novice in cp2k and ab initio molecular dynamics. Recently I tried 
compiling an up-to-date version of CP2K, that is CP2K 4.1, and runned a few 
geometry optimization jobs using it with openmpi-2.0.0. I found that the 
4.1 version with openmpi-2.0.0 is much too slower than the old 2.1 version, 
which means the former takes as much time as 3 or 4 times to finish an 
exactly same jobs using one node with same processor number. 

Here is my compilation configuration for CP2k 4.1 and 2.1.
CP2K 4.1:
using: intel mkl 2015 version, fftw-3.3.4, openmpi-2.0.0, intel fortran 
compiler 2015
CP2k 2.1:
using: intel mkl 2013 version, fftw-3.3.4, openmpi-1.6.5, intel fortran 
compiler 2013

The compilation optimization is referred to the ARCH file 
Linux-ia64-intel.popt provided by cp2k package because out cluster is using 
intel Xeon platform. The compilation is parallelled using 8 cores with 
openmpi.

Can anybody who had experience of using both new and older version of CP2K 
 share me with their performace differences?

THANKS!!!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20170326/eaf68b5a/attachment.htm>