We have tested using c1 instead of c0 but no difference. We don't use logical processors at all. When the problems happens, it doesn't matter what you set the cores for C1/C0, they never get up to speed again without a power cycle/reseat. We believe this to be something related to power. Maybe current limiting.
As I stated yesterday, after a complete chassis power cycle on Tuesday Sept 10, the entire 37 chassis have been outperforming their 2.6GHz ratings flawlessly. I don't know if this is going to be the solution we have been searching to find but it has certainly been a week and a half of some very happy researchers! Thanks, Bill On 09/19/2013 11:32 AM, Christopher Samuel wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 18/09/13 10:49, Douglas O'Flaherty wrote: > >> "Run in C1. C0 over commits unpredictably, then throttles." > I've seen a recommendation in a public Mellanox document of using C1 > not C0 when using hyperthreading/SMT, could be related to this.. > > - -- > Christopher Samuel Senior Systems Administrator > VLSCI - Victorian Life Sciences Computation Initiative > Email: sam...@unimelb.edu.au Phone: +61 (0)3 903 55545 > http://www.vlsci.org.au/ http://twitter.com/vlsci > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.12 (GNU/Linux) > Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ > > iEYEARECAAYFAlI7GPAACgkQO2KABBYQAh9zPQCfeOCdUupjqx7SDeFxQjBWG9NU > FL4AnRYA3zLCNzEVNp0ypiW9KMYp3ohW > =ntfO > -----END PGP SIGNATURE----- > _______________________________________________ > Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf _______________________________________________ Beowulf mailing list, Beowulf@beowulf.org sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
