Dear Peter, I would suggest for checking pw.x the option diagonalization='cg', in this way a conjugate-gradient algorithm is used for finding the ground-state of the KS-Hamiltonian, instead of the 'davidson' algorithm. It will be slower but maybe more stable.
The main difference between the cp.x and pw.x (at gamma-point) codes is that cp.x minimizes directly the total energy avoiding charge-mixing. This could turn into more stability in case of large systems. Paolo ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.