My Community

Hi,

I am calculating Transition State (TS) by Improved Dimer Method. My system has more than 100 atoms. H and OH on a surface. The surface structure is changing after adsorbing H and OH. So, I need to consider all the atoms for vibrational frequency calculation (IBRION=5 or 6) as the first step. Since I have many atoms, the calculation takes very long time on 1 node due to the large number of degrees of freedom. I was trying to parallelize the job over 2 or 4 or more nodes by setting the NCORE, NPAR, or KPAR, but job is terminated every time.

Is it possible to parallelize this kind of run? Below I have attached the type of INCAR I tried for parallelization and the error I am getting. The HPC I am using has 56 cores on each node.

PREC = Normal
NSW = 1
IBRION = 5
ISIF = 2
NCORE = 14
KPAR = 4
SYMPREC = 1E-04
NFREE = 2
POTIM = 0.02
NWRITE = 3

VASP internal routines have requested a change of the k-point set. |
| Unfortunately, this is only possible if NPAR=number of nodes. |
| Please remove the tag NPAR from the INCAR file and restart the |
| calculation. |
| |
| ----> I REFUSE TO CONTINUE WITH THIS SICK JOB ... BYE!!! <----

I tried several combinations between NCORE, NPAR, KPAR, but nothing worked. Could anyone please help me?

Unfortunately, NCORE/=1 is not supported in combination with IBRION=5 or 6.
The only parallelization level you can use is KPAR or the VASP default, which is parallelization over bands.
This will be fixed in an upcoming version of VASP.

How many k-points do you have in your calculation? Can you share your KPOINTS or IBZKPT file?
You can choose KPAR to be close to the number of nodes you want to use and divisible by the number of points in the IBZKPT file.

Hi Henrique,

Thanks for your help. There are 4 KPOINTS (2*2*1). I checked by setting KPAR=4 and submitted the job on 4 nodes. It is at least 3 times faster, I think.

I will update after completion of the job. I appreciate your time.

Thank you.