[CP2K-user] [CP2K:17947] B3LYP performance for QM/MM

[Jürg Hutter]

Hi

Yes, this is about what can be expected. Especially, the use of MOLOPT basis sets
can cause a strong increase in CPU time needed for hybrid functionals.
I would suggest to use the ADMM approximation to speed up your simulation.
With an optimal setup you should be able to get back within a factor of 2-3 to
your BLYP timing.

BTW: you should also change the reference functional in the vdW part to B3LYP.

regards

JH

________________________________________
From: cp2k at googlegroups.com <cp2k at googlegroups.com> on behalf of Max W <winokanmax at gmail.com>
Sent: Thursday, October 27, 2022 4:30 PM
To: cp2k
Subject: [CP2K:17945] B3LYP performance for QM/MM

Hi all,

I have been using the CP2K-Gromacs package on the Archer2 HPC for QM/MM with the BLYP functional, but would like to start using B3LYP. I have run some benchmarks with SPE's for the 2 functionals and my ~30 QM atom system but the performance of B3LYP is around 20 times worse for B3LYP compared to BLYP. Is this to be expected? This seems like a very large cost increase that I did not expect. I am running with 32 tasks each with 4 threads.

I would be very grateful to receive any insight into whether this kind of performance is expected, and/or if there are any obvious performance improvements I should investigate.

I have attached the two input files to this post, but the PDB is too large to share directly. It can be found here: https://app.box.com/s/yyo8cfwsd1rz8n35h02bckyr0cer0lq5

Thank you!

Max

--
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<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/78f2881e-f3d3-4e8f-a135-00688a023ff9n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/78f2881e-f3d3-4e8f-a135-00688a023ff9n%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 on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB07599C7B530AA7A43DC3490F9F329%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.