Sorry!
3) play a bit with ecutfock (ecutfock=2*ecutwfc <--- is generally
safe, even less sometimes) which saves a lot of memory
Giuseppe
Quoting Shoaib :
> Dear Members,
>
>
>
> I am trying a simple HSE calculation but it stop progressing after some time
> where
Dear Shoaib Muhammad
First of all EXX calculations are *far more stable* when you use norm
conserving pseudopotentials. ultrasoft pseudopotentials such as yours
do not even guarantee particular speedup: the bottleneck is on density
cutoff rather than on wavefunction cutoff.
Moreover, EXX
Dear Members,
I am trying a simple HSE calculation but it stop progressing after some time
where output file says "convergence has been achieved in 10 iterations"
without showing any further error. CPU usage still remains high but nothing
is further printed in the output file even if I wait
Dear Paolo,
I got it. Thank you so much!
Best regards,
Hongsheng
2017-02-03 12:13 GMT+01:00 Lorenzo Paulatto
:
> You can also try with less q-points (see nqx1, nqx2, nqx3) as HSE should
> not need an 8x8x8 grid. But I mostly recommend that you start with a
>
You can also try with less q-points (see nqx1, nqx2, nqx3) as HSE should
not need an 8x8x8 grid. But I mostly recommend that you start with a
smaller system to get familiarized.
hth
On 3 February 2017 at 09:37, Hongsheng Liu wrote:
> Dear Paolo,
> Thank you very
Dear Paolo,
Thank you very much. I learn a lot from your reply. So is there any way I
can solve this problem? Or I can only change to another computer with
bigger memory? Thanks a lot!
Best regards,
Hongsheng
2017-02-03 9:29 GMT+01:00 Paolo Giannozzi :
> SCF convergency
SCF convergency is first reached without the exact-exchange part; then
further SCF cycles are performed with the exact-exchange contribution.
In the latter step, ALL Kohn-Sham orbitals for ALL k-points (not just
those in the Irreducible Brillouin Zone) are stored on the real-space
grid. It's a lot
Dear Paolo,
Thank you very much for your reply. I check for the memory. You are
right. During the calculation, only few memory is need. But at the end of
the calculation (just after the convergence of scf), the demand of memory
suddenly increased greatly and the job stopped then. Only 14 atoms
Very likely reason: too much memory required
Paolo
Il 02/feb/2017 05:35 PM, "Hongsheng Liu" ha
scritto:
> Dear all,
> I'm trying to do a HSE scf calculation for bulk magnetite with PWSCF
> v.5.4.0. However, the calculation failed just after the convergence of scf
>
Dear all,
I'm trying to do a HSE scf calculation for bulk magnetite with PWSCF
v.5.4.0. However, the calculation failed just after the convergence of scf
with the error message as below,
'convergence has been achieved in 14 iterations
APPLICATION TERMINATED WITH THE EXIT STRING: Killed
Dear Giuseppe,
Very thanks for your advice!
For the points b) and e), actually I am mainly focused on the wavefunctions in
SCF calculations, is there any change in your recommended values?
Best wishes!
Jiqiang
在 2017-01-05 21:30:51,"Giuseppe Mattioli"
On Thu, Jan 5, 2017 at 2:30 PM, Giuseppe Mattioli <
giuseppe.matti...@ism.cnr.it> wrote:
> i) disk_io = 'none'. Dangerous. It could hinder a proper restart of
> your calculation.
>
it used to, but it should no longer have any adverse effect on restart (as
long as the running code is stopped
Sorry, in my previous post:
b) ecutfock is a very useful handle that reduces the ecutrho fft grid
in the case of exact exchange calculations. You should play around to
find convergence, but in my experience (-->) ecutfock=ecutwfc yields
satisfactory geometry optimizations, and (-->)
Dear Jiqiang Li
EXX calculations *are* expensive. First of all read this file:
your-path-to-QE/Examples/PW/EXX_example/README
This said, you can use a couple of speedup tricks:
a) The Adaptively Compressed Exchange (ACE) algorithm is implemented
in QE 6.0. EXX is computed with ACE if you add
Dear Quantum Espresso users,
When I did a HSE calculation, I found it was very expensive. I am not
sure whether I did it in the right way, and could anyone give me some advice on
the speeding up the calculation?
The following is my input file. Thank you for your consideration!
Best
15 matches
Mail list logo