In my experience there is no big difference between 100 CPU and 50 CPU when you have only one K-point.
2011/7/1 yefare li <[email protected]> > I believe it's ok if you have >100 CPU, I guess you can relax this > system 1~2 weeks, if you set Meshcutoff in 150Ry. > > On Thu, Jun 23, 2011 at 10:11 PM, Alexander Vozny <[email protected]> > wrote: > >> can somebody tell the maximum Si system size with DZP basis set which > >> Siesta can handle? > >> Relax in less than a month? Only Gamma k-point. Using cluster, parallel > >> calculations. > > > > Ruslan, > > I did a graphene flake (i.e. lots of vacuum) with more than 600 atoms, > 300Ry > > cutoff and DZP, on 6 to 18 processors. It was slow, but converged in > about a > > month. > > If you have a bulk Si (i.e. no vacuum) it could be faster (or at least > less > > memory demanding). > > Try to relax using spin-unpolarized calculation first and only then > continue > > with spin, if required. > > Alex. > > >
