[QE-users] [QE-GPU] GTX 1650
Fabrizio Ferrari Ruffino
faferrar at sissa.it
Wed Dec 7 00:04:42 CET 2022
Hi Antimo,
short answer based on my personal experience with the gtx1050 is: yes, speed-up is clearly observable and I would say significant, much lower than V100 or any other hpc GPU though. If you want some numbers I remember a benchmark I did a year ago limited to the V_xc calculation in PW (a few density FFTs per SCF step and some loops on the density grid). For a small test (water molecule) on 1 MPI I obtained about 2x with respect to CPU only on my laptop, while I got 8-9x on a V100 system. From what I could see and hear from other people (not much) acceleration is observable starting from Pascal (cc60) architecture onwards, regarding 'non-hpc' GPUs.
Honestly I never went further than this, but I doubt that single precision mostly based GPUs could provide advantages in terms of performance/consumption. But my technical knowledge on this is very limited, so I don't want to provide any definitive statement.
Best,
Fabrizio
________________________________
From: users <users-bounces at lists.quantum-espresso.org> on behalf of MARRAZZO ANTIMO <ANTIMO.MARRAZZO at units.it>
Sent: Tuesday, December 6, 2022 12:38 PM
To: Quantum ESPRESSO users Forum <users at lists.quantum-espresso.org>
Subject: Re: [QE-users] [QE-GPU] GTX 1650
Dear Fabrizio,
I am very interested in your experience on this, have you ever benchmarked QE PWscf on your “cheap” GTX 1050 versus HPC GPUs such as V100?
Or in other words, is the speedup provided by GTX 1050 significant? Especially considering the relative poor performance these cards have in double precision.
These cards cost an order of magnitude less than top-notch HPC GPUs, so it would be interesting for people to know whether the speed/price is actually potentially interesting.
Thank you!
Best regards,
Antimo
Antimo Marrazzo
Dipartimento di Fisica
Università di Trieste
Strada costiera 11
34151 Trieste
Il giorno 6 dic 2022, alle ore 09:58, Fabrizio Ferrari Ruffino <faferrar at sissa.it> ha scritto:
Hello,
I run sometimes QE on my laptop with the gtx 1050 (mainly for debugging). You just need to download the hpc_sdk toolkit. Of course you are limited to small calculations and I often find troubles when I try to run 2+ MPI on the same GPU. But you can certainly give it a try.
Cheers,
Fabrizio
________________________________
From: users <users-bounces at lists.quantum-espresso.org> on behalf of Marko Boskovic <boskovic at nanosys.ihtm.bg.ac.rs>
Sent: Monday, December 5, 2022 5:44 PM
To: users at lists.quantum-espresso.org <users at lists.quantum-espresso.org>
Subject: [QE-users] [QE-GPU] GTX 1650
Dear all,
Is it possible to run QE-GPU on Nvidia GTX 1650?
Best regards,
Marko Boskovic
University of Belgrade
_______________________________________________
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu<http://www.max-centre.eu/>)
users mailing list users at lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users
_______________________________________________
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
_______________________________________________
Quantum ESPRESSO is supported by MaX (www.max-centre.eu<http://www.max-centre.eu/>)
users mailing list users at lists.quantum-espresso.org<mailto:users at lists.quantum-espresso.org>
https://lists.quantum-espresso.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221206/7746bc14/attachment.html>
More information about the users
mailing list