My Community

I am running a hybrid functional calculation, for which NPAR needs to be set to the number of cores (default). I am running a PBE calculation to preconverge the wave function before running a HSE calculation on top of it. In order to have the same basis set upon restart, I also leave the default NPAR set for the PBE calculation. My system is relatively large, with 1040 electrons. When I run this calculation with many cores (I have tried 256 and 192, allocated in 16-core nodes) I get a segmentation fault - even for the PBE calculation. If I run exactly the same calculation with 64 cores distributed in 4 16-core nodes, the PBE calculation runs with no problems.

Hence my question - is there a hard limit to the number of cores that can be used with the default NPAR? If so, is there a way to estimate which value would be optimal?

There is no limit to the number of cores.

I also have the same question.
At last, I use the default NPAR set,it runs with no problem.
But I get a warnning I can't get the DOSCAR and PDOS for the NPAR/=1.

you can use LORBIT.