My Community

Hi,

I have a problem of parallelization settings when I run VASP (5.3.2) using B3LYP functionals. I've tried different settings of NPAR to get it to work, but none of them is working. Here are several things that I've tried:

1. NPAR = 16 (INCAR), running on 1 node and 16 cores per node: the calculation starts ok but stops at the 5th SCF cycle.

2. NPAR = 32 (INCAR), running on 2 nodes and 16 cores per node: the calculation starts ok but does not enter the 1st SCF cycle.

3. NPAR = 4 (INCAR): running on 2 nodes and 16 cores per node: the calculation does not start, a warning message shows up requiring me to set NPAR = number of cores per node.

4. No NPAR command in INCAR: same result as in case 1.

For all the cases there is a warning message in the OUTCAR file:

====================
For optimal performance we recommend that you set |
| NPAR = 4 - approx SQRT( number of cores) |
| (number of cores/NPAR must be integer) |
| This setting will greatly improve the performance of VASP for DFT. |
| The default NPAR=number of cores might be grossly inefficient |
| on modern multi-core architectures or massively parallel machines. |
| Do your own testing. |
| Unfortunately you need to use the default for hybrid, GW and RPA |
| calculations.
====================

I've tried other things suggested by the manual, such as ALGO = Damped, but none of them could get the calculation to work.

Could you please help me with this? Thank you very much.

The parallelization improves the performance. In no case it causes
that the calculation stops. It seems you are using not correctly compiled
executable.

Hi,

Thank you for your reply. What do you mean by compiled executable? The problem only occurred when I used B3LYP. Everything worked fine when I used other functionals such as PBE.

Best regards,
Andy

Attach your input files.

Hi,

I have problem attaching files, so I just put them here. For pseudopotentials I use the PAW_GGA for Ni and O. This job stops after the 5th cycle. Thank you.

INCAR:
====================
SYSTEM = B3LYP_Test

#NPAR = 16
LPLANE = .TRUE.
LSCALU = .FALSE.
NSIM = 4

ISTART = 0
ICHARG = 2

LHFCALC = .TRUE
GGA = B3
AEXX = 0.2
AGGAX = 0.72
AGGAC = 0.81
ALDAC = 0.19
ISYM = 3
PRECFOCK = A
ALGO = Damped

PREC = Accurate
ENCUT = 500 eV
IBRION = 2
NSW = 100
ISIF = 3
EDIFFG = 1E-3

NELM = 100
EDIFF = 1E-4

ISMEAR = 1
SIGMA = 0.2

ISPIN = 2
MAGMOM = 2.0 -2.0 0.0 0.0
LASPH = .TRUE
LSORBIT = .FALSE
LMAXMIX = 4

LORBIT = 11
====================

POSCAR:
====================
AFM NiO
4.17
1.0 0.5 0.5
0.5 1.0 0.5
0.5 0.5 1.0
2 2
Cartesian
0.0 0.0 0.0
1.0 1.0 1.0
0.5 0.5 0.5
1.5 1.5 1.5
====================

KPOINTS:
====================
k-points
0
gamma
4 4 4
0 0 0
====================

Best regards,
Andy

Hi Andy,

did the calculation really abort or was just no progress visible?

If VASP starts fith HF exchange, you'll get from seconds to many minute(s) execution time per step.

Cheers,

alex

Hi Andy,
with your data I started two calculations.
Lower comutational cost 400 eV, 2 2 2: finished without any problem.
With non-modified input the calculation is running as well.

alex wrote:Hi Andy,

did the calculation really abort or was just no progress visible?

If VASP starts fith HF exchange, you'll get from seconds to many minute(s) execution time per step.

Cheers,

alex

Thank you for your reply.

The calculations finished Iteration 1, Cycle 5 then showed no progress for almost ten hours. Is this what should be expected?

Thank you.

Best regards,
Andy

admin wrote:Hi Andy,
with your data I started two calculations.
Lower comutational cost 400 eV, 2 2 2: finished without any problem.
With non-modified input the calculation is running as well.

Thank you for your reply.

I'll try to run the job again. But if you can run it fine, I guess it means that the inputs are ok? In this case, do you have any causes of this problem in mind?

Best regards,
Andy

Dear admin,

One more question, what version of VASP did you use to run the job? Thank you.

Best regards,
Andy

vasp.5.2.9.

Dear admin,

Thank you for your reply. For your successful job, is there any warning message being shown?

Best regards,
Andy

Just the usual message:
| You enforced a specific xc-type in the INCAR file, |
| a different type was found on the POTCAR file |
| I HOPE YOU KNOW, WHAT YOU ARE DOING |

Dear admin,

Thank you for your reply.

Since you didn't get the warning but I did, could this problem be related to the parallelization settings of the clusters that I am using? Thank you.

Best regards,
Andy

Dear admin,

Following my last post, I want to ask that did you set NPAR = something in the INCAR? I remember even for the previous successful jobs I ran (with other functionals such as PBE), if I didn't set NPAR, the job could run but that warning would show up. Thank you.

Best regards,
Andy