Hi Ryan,
I don't have the numbers to hand right now - Perri, cc'd here should be able to provide them for the standard AMBER benchmarks. If memory serves in native mode for the JAC benchmark each Xeon-Phi card is roughly equivalent to around 6 cores of an 8 core E5-2640v3.
All the best
Ross
> On Feb 5, 2015, at 1:07 PM, Novosielski, Ryan <novosirj.ca.rutgers.edu> wrote:
>
> Any idea how they perform on a system running in native mode (I took that to mean "running on Phi alone, like a GPU.")? We are currently looking at some pretty aggressive pricing on some Phi cards and are weighing options.
>
> ____ *Note: UMDNJ is now Rutgers-Biomedical and Health Sciences*
> || \\UTGERS |---------------------*O*---------------------
> ||_// Biomedical | Ryan Novosielski - Senior Technologist
> || \\ and Health | novosirj.rutgers.edu<mailto:novosirj.rutgers.edu>- 973/972.0922 (2x0922)
> || \\ Sciences | OIRT/High Perf & Res Comp - MSB C630, Newark
> `'
>
> On Feb 5, 2015, at 15:57, Ross Walker <ross.rosswalker.co.uk<mailto:ross.rosswalker.co.uk>> wrote:
>
> Hi Ryan,
>
> Perri is in the process of writing a paper to describe the Xeon Phi work and will then put together a webpage on it which will have performance numbers etc. For now you get roughly about a 10 to 15% boost in performance, for >400K atoms, by adding a Xeon Phi to a couple of 8 core CPUs.
>
> All the best
> Ross
>
> On Feb 5, 2015, at 12:45 PM, Ryan Novosielski <novosirj.ca.rutgers.edu<mailto:novosirj.ca.rutgers.edu>> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Thanks for your reply.
>
> The one is basically a snippet of the documentation that I had read,
> and the other is the mention of MIC/Phi support, which I knew from the
> documentation. Did I miss benchmarks or at least a ballpark of
> performance compared to a CPU or GPU?
>
> Thanks!
>
> On 02/04/2015 07:42 PM, Ross Walker wrote:
> Hi Ryan,
>
> Yes - please see update.9 and the following updated documentation:
> http://ambermd.org/MICv2_Amber14.pdf
>
> http://ambermd.org/bugfixes14.html
>
> All the best Ross
>
> On Feb 4, 2015, at 3:48 PM, Novosielski, Ryan
> <novosirj.ca.rutgers.edu<mailto:novosirj.ca.rutgers.edu>> wrote:
>
> Has there been any update on Xeon Phi support in Amber? I have
> read the section of the manual that explains how it is
> configured, and the mention of adding support on the website. But
> that same section mentions that benchmarks are forthcoming. I'm
> wondering if anyone has a rough idea of how they compare in
> performance to the other stuff that is out there (Kepler, CPU,
> etc.).
>
> ____ *Note: UMDNJ is now Rutgers-Biomedical and Health Sciences*
> || \\UTGERS |---------------------*O*---------------------
> ||_// Biomedical | Ryan Novosielski - Senior Technologist || \\
> and Health | novosirj.rutgers.edu<mailto:novosirj.rutgers.edu><mailto:novosirj.rutgers.edu>-
> 973/972.0922 (2x0922) || \\ Sciences | OIRT/High Perf & Res
> Comp - MSB C630, Newark `'
> _______________________________________________ AMBER mailing
> list AMBER.ambermd.org<mailto:AMBER.ambermd.org>
> http://lists.ambermd.org/mailman/listinfo/amber
>
>
> _______________________________________________ AMBER mailing list
> AMBER.ambermd.org<mailto:AMBER.ambermd.org> http://lists.ambermd.org/mailman/listinfo/amber
>
>
> - --
> ____ *Note: UMDNJ is now Rutgers-Biomedical and Health Sciences*
> || \\UTGERS |---------------------*O*---------------------
> ||_// Biomedical | Ryan Novosielski - Senior Technologist
> || \\ and Health | novosirj.rutgers.edu<mailto:novosirj.rutgers.edu> - 973/972.0922 (2x0922)
> || \\ Sciences | OIRT/High Perf & Res Comp - MSB C630, Newark
> `'
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1
>
> iEYEARECAAYFAlTT1nMACgkQmb+gadEcsb7+3QCeMVVbxaMvJXRJd/JNcBP7Le3t
> 1qwAn2xF4hY8eBO9yz4xmwfe/33tNDEx
> =b+uG
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org<mailto:AMBER.ambermd.org>
> http://lists.ambermd.org/mailman/listinfo/amber
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org<mailto:AMBER.ambermd.org>
> http://lists.ambermd.org/mailman/listinfo/amber
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 05 2015 - 13:30:06 PST
