UT theoretical chemistry code forum

Hello everyone:
I met a problem that everytime I run a NEB job with 8 images on 4 nodes(there are 16 cores each nodes), the vasp would terminated and shows the error that: "APPLICATION TERMINATED WITH THE EXIT STRING: Segmentation fault (signal 11)", If I increase the nodes to 8 nodes, it will work well, but it will require much more cores and I can hardly have the chance to get my job calculated in this way because we are using the PBS quene system to distribute calculating resource.
I have searched on the Internet and found that it has something to do with the RAM given to VASP, Is anybody knows the detail, could someone give me some tips to solve the problem?
Thank you.
Best Wishes.

First, this is posted on a forum related to the EON code. Is this really an EON calculation, or standard VASP-NEB?

I assume it is the latter. If so, I think your assessment is correct, that you need to run each image on a sufficient number of cores to distribute the memory per core.

To help with the problem, take a look at your calculation and make sure that it needs as much memory as you are asking for. Do not use excessive numbers of k-points for an initial NEB calculation; keep the cutoff reasonable; use soft potentials where possible; only use spin polarization if needed; and use precision normal. Once you have located the saddle, you can switch to a min-mode following calculations (e.g. a dimer calculation set up with the neb2dim.pl script) and then refine the saddle with a single image using whatever accurate parameters that are appropriate.