[Pw_forum] Too much memory required

[Pablo Álvarez Zapatero]

Hi everyone,


I am traying to perform a scf calculation for a 756 atoms supercell nanoribbon (for a phonon -phonopy- calculation). The problem is that it using the next input:


&control
 calculation = 'scf'
 prefix='grapheneph',
 pseudo_dir = './',
 outdir='./saleph62/'
/
&system
 ibrav = 0,
 nat = 756,
 ntyp = 1,
 occupations = 'smearing',
 smearing = 'gaussian',
 degauss = 0.01,
 ecutwfc = 60.D0,
 ecutrho = 500.D0,
 /
&electrons
    conv_thr = 1.D-8,
    diagonalization = 'cg',
    mixing_mode = 'plain',
/
K_POINTS gamma
!    ibrav = 0, nat = 756, ntyp = 1
CELL_PARAMETERS bohr
   89.4600000000000080    0.0000000000000000    0.0000000000000000
    0.0000000000000000   48.4470000000000027    0.0000000000000000
    0.0000000000000000    0.0000000000000000   10.0000000000000000
ATOMIC_SPECIES
  C   12.01070   C.pbe-rrkjus.UPF
ATOMIC_POSITIONS crystal

[756 Carbon atoms...]


Requires Estimated max dynamical RAM per process >  64279.60Mb of memory, which results in a memory allocation error, since I ''only'' have 65Gb of memory in my computer.

I have already read across this forum and pages from the QE manual, where it suggests using a gamma k-point for instance. But i do not know if this huge amout of memory required is normal, or is due to some absurd input option I am using.


Using electron_maxstep = 50 tag does not lead to a memory error, although it is quite slow.


Solving with this same input for a 1x1x1 supercell (78 atoms), it works perfectly.


Any ideas why I am having this problem? Should I use other input options for this system?


Thank you for your time and patience.

Pablo.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170504/c6d78ab9/attachment.html>