[QE-users] [SPAM] GPU-accelerated pw.x error: "cusolverDnZhegvdx failed"

[Nagy Gergely Norbert]

Dear Paolo,

Thank you so much for your prompt reply and running the test. I have followed your suggestion and submitted a calculation with only two CPU cores and the "-nk 2" option. Unfortunately, the result was exactly the same. It appears the issue is on our end.

We will request the installation of the latest QE version on our HPC and run tests with it. Hopefully that will fix the issue.

Thank you again for your assistance.

Best regards,
Gergely Nagy



Gergely Nagy  –  ELI-ALPS

________________________________
From: Paolo Giannozzi <paolo.giannozzi at uniud.it>
Sent: Tuesday, 17 September 2024 8:24 am
To: Quantum ESPRESSO users Forum <users at lists.quantum-espresso.org>
Cc: Nagy Gergely Norbert <Gergely.Nagy at eli-alps.hu>
Subject: Re: [QE-users] [SPAM] GPU-accelerated pw.x error: "cusolverDnZhegvdx failed"

[You don't often get email from paolo.giannozzi at uniud.it. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

It may or may not be related: if you have 2 GPUs, you should run 2 MPI
processes, preferably with k-point parallelization. Something like:
mpirun -np 2 pw.x -nk 2 ...

Note the message:

>       GPU acceleration is ACTIVE.
>
>       Message from routine print_cuda_info:
>       High GPU oversubscription detected. Are you sure this is what you
> want?

Paolo
--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine Italy, +39-0432-558216
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20240917/38000e29/attachment.html>