[CP2K:8645] Comparison between CP2K and Quantum Espresso

[hut... at chem.uzh.ch]

Hi

without additional information (e.g. output files or complete input)
I will have to do guessing.

There are two parts to this calculation
- time per SCF step
- number of SCF steps needed for convergence

SCF convergence: you start from an atomic guess (CP2K) and a random
wfn (QE). Although the convergence asked for is the same (1.e-7)
this means different things in the two codes (check change in energy
for the last step). For CP2K you can optimize the settings for the
convergence, but as you do start from scratch only once the more 
important part is what is the timing per step.
(You probably also don't need CHOLESKY OFF?, 
 you use the most expensive PRECONDITIONER,
 DIIS might not be good to start with ...)

Your settings are ok, but rather tight in many respect, e.g. large and
diffuse basis set, high cutoff. Non orthorhombic cell is also not
optimal for CP2K.
So this is a very expensive calculation in CP2K standards.

regards

Juerg

--------------------------------------------------------------
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: tfran... at gmail.com
Sent by: cp... at googlegroups.com
Date: 02/05/2017 08:06PM
Subject: [CP2K:8645] Comparison between CP2K and Quantum Espresso

Dear all,I am currently testing my new machine, which is a cluster 2 Supermicro 2048U RT4 nodes with 4x2.6GHZ Intel Broadwell 10Core CPUs.The point is that in order to check the scalability and the performances of the machine with 40 processors, I run several single point calculations of the same structure both with CP2K version 4.1 ssmp and Quantum Espresso 5.1.1 It turns out that for the time spent by CP2K to converge a single point calculation with 40 processors, for a system of 256 atoms, was of the order of 13316.63 seconds (taken from the output in the TIMING BLOCK in the end of the output file [TOTAL TIME AVERAGE); whereas for the QE, the same calculation, with the same system ect. took just 1845.76 seconds. So, because we strongly want to work with CP2K, also for running some dynamics and exploit its scalability, may i ask you to kindly check if the input file of the CP2K I used is properly set? I am more expert in QE rather than CP2K, so I might have miss something essential.
For matter of comparison I attached the two input files used, one for the CP2K and the other of the QE.
Thanks in advance for the kind help!
                
        
        
                
                        
                                
                                        


                                
                        
                
                
        
        
                
                        
                                
                                        Best Regards,Tommaso 



                                
                        
                



-- 

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 cp2k+uns... at googlegroups.com.

To post to this group, send email to cp... at googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

For more options, visit https://groups.google.com/d/optout.



[attachment "cp2k.inp" removed by Jürg Hutter/at/UZH]
[attachment "qe.inp" removed by Jürg Hutter/at/UZH]