memory error

[divanshu_gupta]

Hi VASP Community,
I am running a hybrid functional calculation for a cell with 36 atoms. I am doing a single point scf calculation with hybrid functional. However, I am getting out of memory error or if the job starts, it only does 4 iterations/steps, gets stuck, and the job finishes when the wall time is reached. I am attaching the VASP files for your reference.

I also tried playing around with different NPAR/NCORE/KPAR tags but it did not work for me.

I also tried changing the hardware specification that I used, especially with different numbers of nodes(I have tried with 2,4,6), cores/node=48, mempercpu=2gb(I tried with 4gb as well).

Any suggestion would be helpful.
Divanshu

Hybrid_scf_calculation.zip

[andreas.singraber]

Hello!

I believe your INCAR and KPOINTS settings result in a very expensive calculation regarding both, memory consumption and computation time. In your INCAR file you set

Code: Select all

ENCUT = 850

which seems quite large compared to the numbers in the POTCAR file:

Code: Select all

ENMAX  =  400.000; ENMIN  =  300.000 eV
ENMAX  =  250.000; ENMIN  =  200.000 eV

Also, your KPOINTS file generates an 18 x 18 x 1 mesh which looks expensive. I would recommend to start with much lower settings, e.g., ENCUT = 400 and 4 x 4 x 1 and measure the timing and memory consumption. Then, gradually increase the numbers until you can estimate at which settings you can still afford the calculation and convergence is good enough.

You mentioned that sometimes your jobs actually run but seem to be stuck after 4 iterations. In this case, VASP is actually not stuck but the first 4 iterations are done only with PBE and no Hartree-Fock is used (see at the bottom of NELMDL). Then, at the fifth iteration Hartree-Fock is just so expensive (because of your settings) that it does not finish within the allowed wall time.

Hope this helps!

All the best,
Andreas Singraber