Bench.Hg and Benchmark results VASP

Message

dnumb · #1 Post by **dnumb** » Thu Feb 18, 2010 2:37 pm

Hi there,

I have ran the Benchmark and Bench.Hg files from the VASP server as the manual says and got results that differ from those expected in the reference output files (OSZICAR.ref).
For example, I got Energy difference of ~0.3kJ/mol for the Benchmark calculation and 7kJ/mol for the Bench.Hg calculation. I think the first isn't significant but the second is of the order of polimorphic energy differences and is significant considering these are benchmarks and are run as single-point energy calculations where the atomic positions are not optimized.
I was therefore wondering if anybody got this before or could share some light on it please.

Cheers

David

fanghz · #2 Post by **fanghz** » Tue Mar 02, 2010 8:40 am

The differences are both too large! One of the possible reasons might be the wrong installation of VASP math libarary, which I met before. If you are playing with the parrallel version, I recommend switching to the serial version if possible, then check the difference again.

dnumb · #3 Post by **dnumb** » Tue Mar 02, 2010 11:39 am

Hi fanghz,

Math libs... but I didn't have too many problems compiling... Anyways I'll do that and see!

babaleo · #4 Post by **babaleo** » Tue Mar 02, 2010 9:02 pm

Hi, this is my results in comparison with .ref file. I am not sure I should worry about the difference.
[babaleo]$diff OSZICAR OSZICAR.ref
2,15c2,15
< RMM: 1 -0.514507058760E+05 -0.51451E+05 -0.13177E+05 316 0.780E+02
< RMM: 2 -0.527604338585E+05 -0.13097E+04 -0.23675E+04 316 0.234E+02
< RMM: 3 -0.529743353769E+05 -0.21390E+03 -0.41254E+03 316 0.116E+02
< RMM: 4 -0.531145169974E+05 -0.14018E+03 -0.15769E+03 316 0.784E+01
< RMM: 5 -0.531789029672E+05 -0.64386E+02 -0.67142E+02 316 0.452E+01
< RMM: 6 -0.532264453365E+05 -0.47542E+02 -0.47991E+02 720 0.309E+01
< RMM: 7 -0.532330334404E+05 -0.65881E+01 -0.94371E+01 762 0.919E+00 0.871E+00
< RMM: 8 -0.532322794399E+05 0.75400E+00 -0.37182E+01 697 0.816E+00 0.265E+00
< RMM: 9 -0.532327283041E+05 -0.44886E+00 -0.88476E+00 702 0.383E+00 0.129E+00
< RMM: 10 -0.532327148451E+05 0.13459E-01 -0.69686E-01 695 0.120E+00 0.550E-01
< RMM: 11 -0.532327089538E+05 0.58913E-02 -0.18550E-01 693 0.501E-01 0.247E-01
< RMM: 12 -0.532327075118E+05 0.14420E-02 -0.34613E-02 691 0.226E-01 0.756E-02
< RMM: 13 -0.532327075990E+05 -0.87219E-04 -0.65477E-03 688 0.823E-02
< 1 F= -.53232708E+05 E0= -.53232710E+05 d E =0.749687E-02
---
> RMM: 1 -0.514560573315E+05 -0.51456E+05 -0.13169E+05 316 0.781E+02
> RMM: 2 -0.527656755389E+05 -0.13096E+04 -0.23552E+04 316 0.234E+02
> RMM: 3 -0.529900570826E+05 -0.22438E+03 -0.41261E+03 316 0.116E+02
> RMM: 4 -0.531239491511E+05 -0.13389E+03 -0.15259E+03 316 0.781E+01
> RMM: 5 -0.532150078487E+05 -0.91059E+02 -0.94443E+02 741 0.443E+01
> RMM: 6 -0.532325412870E+05 -0.17533E+02 -0.20957E+02 898 0.136E+01
> RMM: 7 -0.532353272484E+05 -0.27860E+01 -0.42986E+01 721 0.429E+00 0.845E+00
> RMM: 8 -0.532326964177E+05 0.26308E+01 -0.15892E+01 697 0.523E+00 0.235E+00
> RMM: 9 -0.532326234710E+05 0.72947E-01 -0.26559E+00 702 0.257E+00 0.657E-01
> RMM: 10 -0.532326396761E+05 -0.16205E-01 -0.32064E-01 694 0.888E-01 0.358E-01
> RMM: 11 -0.532326387202E+05 0.95591E-03 -0.49341E-02 690 0.364E-01 0.898E-02
> RMM: 12 -0.532326385045E+05 0.21567E-03 -0.65914E-03 692 0.968E-02 0.548E-02
> RMM: 13 -0.532326384716E+05 0.32901E-04 -0.15794E-03 599 0.566E-02
> 1 F= -.53232638E+05 E0= -.53232641E+05 d E =0.638724E-02

alex · #5 Post by **alex** » Wed Mar 03, 2010 8:01 am

Hi, if you decrease EDIFF you shoud get a better converged value, However, it'll take more cycles. I'd not worry about the differences, because they are smaller than the convergence criterium ...

Cheers,

alex

dnumb · #6 Post by **dnumb** » Wed Mar 03, 2010 3:31 pm

Hi Alex,

Those results are for Benchmark or Bench.Hg?

I got exactly the same for the latter files, which gives an Ediff=0.07eV. Regarding the convergence criterion isn't it 0.01, i.e. this is actually 7 times bigger?

alex · #7 Post by **alex** » Wed Mar 03, 2010 4:16 pm

I was refering to EDIFF in the INCAR file. However, the step to step convergence is smaller than 10**-4, e.g. 0.32901E-04 in the second OSZICAR. The original Hg bench was probably calculated on a Pentium I or so. So, I would not worry about ...

Cheers,

alex

babaleo · #8 Post by **babaleo** » Thu Mar 04, 2010 2:40 am

Thanks, Alex, now I can trust my compilation. Since the benchmark run only took 60 sec (intel i7, fedora ), but the .ref file has ~450 sec, so I was not sure if I compiled the code successfully. If the .ref file is from P1 machine as you said, then the CPU time is also reasonable now. Thanks.

#9 Post by **admin** » Wed Mar 24, 2010 9:17 am

calculated with vasp.4.6.35, the total energy of the Hg benchmark is
1 F= -.53232708E+05 E0= -.53232710E+05 d E =0.749687E-02
(difference of 0.0135 kJ/mol compared to the reference files, generated with vasp.4.4.4)
the results of the benchmark calculated with vasp.5..2.2 are identical to the 4.6.35 result up to the last digit

dnumb · #10 Post by **dnumb** » Wed Mar 24, 2010 3:08 pm

Thanks admin!!!
This means you got the same as me

But how did I calculated it to be ~7kJ/mol and you got 0.0135kJ/mol?
The energy difference from the OSZICAR and OSZICAR.ref is 70meV and 1eV = 96.4852912 kJ/mol or is this wrong??

#11 Post by **admin** » Thu Mar 25, 2010 11:30 am

sorry, it is 0.135, not 0.0135 (I must have divided the energy difference by 500 atoms instead of 50 in a hurry). please note that the unit cell contains 50 Hg atoms, therefore if you give the difference in kJ/mol, the number of atoms in the cell has to be accounted for