In a supercom with 40 cores/node, normally when I run VASP for a large system with a few hundreds atoms, I set NCORE=20 expecting that 20 cores treat 1 band inside node. Is it OK to apply the same INCAR setting for NEB calculation? When I run with 8 images, I found that all the 8 directories run at the same time. Then, what is happening here? 20 cores for 1 band is applied to 8 calculations at the same time, or do I need to change NCORE if I expect the same performace?
which NEB used in your work. VASP's own neb or VTST?
VTST is better than vasp's own neb.
Haeju Choi, I have run NEB in VASP. Thank you for information. I will check VTST.
I have found this site, https://www.vasp.at/wiki/index.php/Category:Parallelization, where Nnode would be multiple of IMAGES. For example, if I use 20 Nnode, IMAGE=10 where each images run 2 Node. Then NCORE would be set with 80 cores in 2 Nodes such as NCORE=20 or 10, etc.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs