Comparison of NEB calculation durations

[ArtKa]

Does anybody compared the duration of the NEB calculation in VASP with some other packages?

I compared the duration of NEB calculations in VASP and CP2k for the same compound (KAlO2; number of IMAGES is the same for both calculations = 6; energy cutoff is also the same; GGA-PBE potentials). For VASP calculation I used 144 cores and for CP2k - 48 cores. And my results shown that NEB in CP2k up to 5 times faster than in VASP even at different number of cores.
Can anybody confirm that VASP NEB is slower than other NEB implementations?


My INCAR:

general:
System = KAlO2
ISTART = 0
IBRION = 3
SMASS = 2
ENCUT = 1500
ISPIN = 1
ALGO = VeryFast
LREAL = Auto
PREC = High
SIGMA = 0.2
ISMEAR = 1
NSW = 300
EDIFFG = -0.01
ISIF = 2
SPRING=-5
IMAGES=6
POTIM=0.1

[alex]

Hello ArtKa,

both codes differ heavily wrt. employed physics, e.g. VASP uses PAW and plane waves as basis set, CP2K an LCAO approach.

You would have to go to a well converged limit of a certain/interesting property and compare again ...

Hth,

alex

[ArtKa]

Hello Alex!

Thanks for your reply!
Of course I understand that underlying physics of the codes is different.
But, anyway, I am interested in activation energy, and for both codes the value of energy is the same (also energy profile along the migration path is the same). If it is so, it is not important which physics is behind the codes (from computational point of view).

[admin]

Note that activation energy can be obtained in a single point calculation
http://cms.mpi.univie.ac.at/vasp/vasp/I ... ethod.html
Demanding NEB calculation is not necessary.

[ArtKa]

Dear admin
Thank you so much! I did not hear about dimer method before. I'll try to use it.
But,back to the topic theme, is there any information about comparison of NEB's speed in different packages?

[admin]

No, your evaluation is the first published here.