I noticed the convergence for some of my non-collinear calculations with spin-orbit coupling will improve (i.e. requires less than half as many scf cycles to converge) if I double the number of processors used for the calculation. What could be the reason for this?
Page: [1]
Re: Better convergence for larger number of processors in case non-collinear calculations with SOC( No.1 )
Can you send me input file and launch script? If it is not very time expensive, I'll repeat calculations. So we can check at least how effect depend on "computer".
This thread is locked.Only browsing is available.
Better convergence for larger number of processors in case non-collinear calculations with SOC
Re: Better convergence for larger number of processors in case non-collinear calculations with SOC