[CP2K-user] [CP2K:22167] Re: Optimization inputs PDOS with HSE06 level of theory and AI-MD

[Frederick Stein]

Dear Lorenzo,
Your timing report reveals that grid operations take about 50 % of your 
runtime. As already suggested by Marcella, you may try to increase 
EPS_DEFAULT. In addition, you may try to reduce CUTOFF or increase the 
number of grids (NGRIDS). Please check whether the accuracy matches your 
expectations. Finally, you may also try to use 2 OpenMP threads per MPI 
rank and reduce the number of MPI ranks accordingly to decrease 
communication.
Best regards,
Frederick

Lorenzo Lagasco schrieb am Montag, 6. April 2026 um 12:23:18 UTC+2:

> Good morning,
> my bad, I've now attached both input and output files (with final timing 
> report). In this case, I repeated the calculation using 2 nodes with 40 
> processors each.
> Thanks for your patience.
>
> Best regards
> Lorenzo Lagasco
>
> Il giorno lun 6 apr 2026 alle ore 11:24 Marcella Iannuzzi <
> marci... at gmail.com> ha scritto:
>
>>
>> Hi,
>>
>> As Frederick mentioned, for a proper analysis one would need the timing 
>> report at the end of the output. 
>> You could speed up the calculation by increasing  EPS_DEFAULT (10^-10) 
>> I would also set CALCULATE_C9_TERM to false. 
>> Kind regards
>> Marcella
>>
>> On Monday, April 6, 2026 at 8:42:29 AM UTC+2 lagasco... at gmail.com wrote:
>>
>>> Good evening,
>>> here attached input and output of my ai-md calculation. In this case, 4 
>>> nodes with 16 processors have been used (the ones with 52 priocessors are 
>>> non available at the moment). Moreover, I'm using the version 2025.1 on a 
>>> cluster (CPU model name Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz).
>>>
>>> Best regards
>>> Lorenzo Lagasco
>>>
>>> Il giorno sab 4 apr 2026 alle ore 20:38 Frederick Stein <f.s... at hzdr.de> 
>>> ha scritto:
>>>
>>>> Dear Lorenzo,
>>>> Sorry, I forgot about your second question. Others may correct me if 
>>>> necessary. I can only guess. Can you run 2-3 MD steps and send us the 
>>>> timing report at the end of the CP2K output file? Do you run with 52 cores 
>>>> or do you use less? What hardware are you running on and which software 
>>>> stack do you use? Which version of CP2K are you talking about?
>>>> Usually, standard-DFT-calculations are most performant if the number of 
>>>> MPI ranks is a square number and you use at most 2 OpenMP threads per MPI 
>>>> rank.
>>>> Best,
>>>> Frederick
>>>>
>>>> Lorenzo Lagasco schrieb am Samstag, 4. April 2026 um 19:25:02 UTC+2:
>>>>
>>>>> Good afternoon,
>>>>> thanks for the information!  In the meantime, I’d like to take the 
>>>>> opportunity to ask if you also have any advice regarding my second 
>>>>> question. 
>>>>> Best regards,
>>>>> Lorenzo Lagasco
>>>>>
>>>>> Il giorno sab 4 apr 2026 alle ore 11:18 Frederick Stein <
>>>>> f.s... at hzdr.de> ha scritto:
>>>>>
>>>>>> Dear Lorenzo,
>>>>>> It is highly recommended to perform hybrid calculations using the 
>>>>>> ADMM approach (
>>>>>> https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/AUXILIARY_DENSITY_MATRIX_METHOD.html
>>>>>>  and https://manual.cp2k.org/trunk/methods/dft/hartree-fock/admm.html). 
>>>>>> You will need to choose suitable fitting basis sets (BASIS AUX_FIT <basis 
>>>>>> set name>) in the respective kind sections usually from BASIS_ADMM or 
>>>>>> BASIS_ADMM_UZH. This should accelerate your calculations significantly.
>>>>>> Regards,
>>>>>> Frederick
>>>>>>
>>>>>> Lorenzo Lagasco schrieb am Freitag, 3. April 2026 um 06:19:22 UTC+2:
>>>>>>
>>>>>>> Hello everyone,
>>>>>>> I'm a new cp2k user and I have a couple of questions, not strictly 
>>>>>>> related to each other:
>>>>>>>
>>>>>>>    1. I have written an input file for a band structure calculation 
>>>>>>>    of a system consisting of delafossite CuAlO₂ with an organic dye (coumarin 
>>>>>>>    343) anchored on its surface, using the HSE06 level of theory. However, the 
>>>>>>>    calculation appears to be unusually slow (using 4 nodes with 52 processors 
>>>>>>>    each, after about 2 hours it has not even completed the first SCF step). I 
>>>>>>>    would like to ask whether it would be possible to get feedback on how to 
>>>>>>>    optimize the input file, or if there might be some incorrectly set 
>>>>>>>    parameters. I have attached both the input file and the xyz file of the 
>>>>>>>    system geometry. 
>>>>>>>    2. At the same time, I am working on NVT CP2K dynamics at the 
>>>>>>>    PBE+D3 level of theory for aluminosilicate nanosheets in water of different 
>>>>>>>    sizes (the simulation box ranges from about 719 to 900 atoms). Currently, 
>>>>>>>    the simulations proceed at a speed of about 25 seconds per MD step (using 1 
>>>>>>>    node with 52 processors for each simulation). Since I need a trajectory of 
>>>>>>>    60 ps with a timestep of 0.5 fs, I would like to ask whether it is possible 
>>>>>>>    to improve the computational performance. I am also attaching the input 
>>>>>>>    file in this case md-GPW.inp
>>>>>>>    
>>>>>>>    Thanks for the help!
>>>>>>>
>>>>>>> -- 
>>>>>> You received this message because you are subscribed to the Google 
>>>>>> Groups "cp2k" group.
>>>>>> To unsubscribe from this group and stop receiving emails from it, 
>>>>>> send an email to cp2k+uns... at googlegroups.com.
>>>>>> To view this discussion visit 
>>>>>> https://groups.google.com/d/msgid/cp2k/7634d177-82ae-445f-aa43-9068ad1d9392n%40googlegroups.com 
>>>>>> <https://groups.google.com/d/msgid/cp2k/7634d177-82ae-445f-aa43-9068ad1d9392n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>>>> .
>>>>>>
>>>>> -- 
>>>> You received this message because you are subscribed to the Google 
>>>> Groups "cp2k" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send 
>>>> an email to cp2k+uns... at googlegroups.com.
>>>>
>>> To view this discussion visit 
>>>> https://groups.google.com/d/msgid/cp2k/fc2e811b-1ae9-4572-a3d2-1c1e14a2d046n%40googlegroups.com 
>>>> <https://groups.google.com/d/msgid/cp2k/fc2e811b-1ae9-4572-a3d2-1c1e14a2d046n%40googlegroups.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to cp2k+uns... at googlegroups.com.
>>
> To view this discussion visit 
>> https://groups.google.com/d/msgid/cp2k/37acddf8-506b-4a3b-a4cd-e2fa03010058n%40googlegroups.com 
>> <https://groups.google.com/d/msgid/cp2k/37acddf8-506b-4a3b-a4cd-e2fa03010058n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/82d23fda-3a28-4a80-bec7-7bb7ac08d19dn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260406/b3adccfc/attachment.htm>