[QE-users] How to quess computational expenses for small organic molecular crystals (PBE0 functional) ...

[Michal Husak]

Hi

I need to get a guess for computational cost  for small organic 
crystalline system + PBE0 functional ...
I had run multiple scf of different similar system on identical HW 
(24 core node) ...

 From my understanding, the computational time should be 
approximately proportional to N^3 ....
The N should be number of electrons handled by the calculations ...

The question is what should be used as N:

1) I belive it should be the number of valance electrons calculated 
for asymmetric unit cell content.
e.g. for crystalline urea (C N2 O1 H4) crystallizing in  P -4 21 m (4 
molecule in unit cell)
the N shoudl be        1*4+2*5+1*6+4*1 =  24.
It should be not 4*24 because the code utilize symmetry not to 
calculate symmetrically dependent molecules ...
The N should not include non-valence electrons, become they are 
effectively handed by pseudo-potentials ...

2) I belive the k-points should be include in form
Nkpoits * N^3  to get a complexity guess, because each  k-point is 
just an additional calculation of a N^3 system ...

Unfortunately my result show I am wrong ...
I get up to 50x discrepancy in calculation cost than guess from 1) 
and 2) assumption ...
The starting structures should be  OK according geometry (good X-ray result).
I perform only single point energy calculation ...

Any related idea ?

Michal