[CP2K-user] [CP2K:15056] The calculation speed of SCAN

hut... at chem.uzh.ch hut... at chem.uzh.ch
Thu Apr 1 12:25:40 UTC 2021


Hi

without seeing the timings I will have to guess:
the increase is coming from more time needed for
1) libxc
2) FFT
3) collocate/integrate
 
you can try to set NGRIDS in
CP2K_INPUT / FORCE_EVAL / DFT / MGRID
to a higher value (e.g. 6). That should help with 3)

Another solution might be to work with GAPW and a lower cutoff,
but that needs testing.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter                         Phone : ++41 44 635 4491
Institut für Chemie C                FAX   : ++41 44 635 6838
Universität Zürich                   E-mail: hut... at chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp... at googlegroups.com wrote: -----
To: "cp2k" <cp... at googlegroups.com>
From: "zh... at gmail.com" 
Sent by: cp... at googlegroups.com
Date: 04/01/2021 04:59AM
Subject: [CP2K:15056] The calculation speed of SCAN

Dear all,

Now I am calculation bulk water using SCAN (BOMD). According to the literature, the energy cutoff of SCAN is 1200 Ry, which is much higher than PBE+D3 (500 Ry).

My system is 64 water molecules in a cubic box. 
Using 24 CPUs, the speed of SCAN is ~ 50 s/step that is much slower than PBE+D3 (~ 9 s/step). I have experience on SCAN using VASP that won’t cost much more time than PBE.

Can anyone share the experience on SCAN, especially for the calculation speed and acceleration of calculation?

The attachment is my input file.

Best,
Ke


  
  -- 
 You received this message because you are subscribed to the Google Groups "cp2k" group.
 To unsubscribe from this group and stop receiving emails from it, send an email to cp... at googlegroups.com.
 To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1a220e95-5c0d-4969-98f1-7644de465301n%40googlegroups.com.
 

[attachment "water.inp" removed by Jürg Hutter/at/UZH]



More information about the CP2K-user mailing list