Two same calculations were run from different accounts in hpc-cluster, each on 2 nodes containing in total 48 cores. But in once case I am getting my calculations complete while in other case it is getting stopped in between. The incomplete calculation has been re-run but the problem still persists. All the input files required are same. What could be the possible reason for this ??
Hello Prabhleen Kaur,
What is written in the log when the simulation stopped on the hcp on which you have trouble ? It could be very helpful.
I would say that it could come from the difference of module charged when you are submitting the job.
So the list of modules charged would be also helpful and someone might have already experienced the same trouble using the same modules.
Maybe you could also search/ask your question on the VASP dedicated forum (https://cms.mpi.univie.ac.at/vasp-forum/forum.php)
Thank You Matthieu David for your response. Following is the error I am getting in the log
ERROR FEXCP: supplied Exchange-correletion table
is too small, maximal index : 4077
But then this I am getting only in one case while other job is completed.
I did not get such error in VASP, but it seems to be a convergence issue, which is common problem in numerical methods. The solution to avoid this error is 1) using a slightly different initial structure or 2) manipulating the exchange correlation factors. These factors are using a fraction of calculated properties (energy, forces, etc.) from previously converged step. Reducing the fraction (close to 0) value makes overall convergence faster and increasing it (close to 1) makes the step convergence faster.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running