[gpfsug-discuss] tsgskkm stuck---> what about AMD epyc support in GPFS?
Simon Thompson
S.J.Thompson at bham.ac.uk
Fri Sep 4 10:02:29 BST 2020
Of course, you might also be interested in our upcoming Webinar on 22nd September (which I haven't advertised yet):
https://www.spectrumscaleug.org/event/ssugdigital-deep-dive-in-spectrum-scale-core/
... This presentation will discuss selected improvements in Spectrum V5, focusing on improvements for inode management, VCPU scaling and considerations for NUMA.
Simon
On 04/09/2020, 08:56, "gpfsug-discuss-bounces at spectrumscale.org on behalf of Jonathan Buzzard" <gpfsug-discuss-bounces at spectrumscale.org on behalf of jonathan.buzzard at strath.ac.uk> wrote:
On 02/09/2020 23:28, Andrew Beattie wrote:
> Giovanni, I have clients in Australia that are running AMD ROME
> processors in their Visualisation nodes connected to scale 5.0.4
> clusters with no issues. Spectrum Scale doesn't differentiate between
> x86 processor technologies -- it only looks at x86_64 (OS support
> more than anything else)
While true bear in mind their are limits on the number of cores that it
might be quite easy to pass on a high end multi CPU AMD machine :-)
See question 5.3
https://www.ibm.com/support/knowledgecenter/STXKQY/gpfsclustersfaq.pdf
192 is the largest tested limit for the number of cores and there is a
hard limit at 1536 cores.
From memory these limits are lower in older versions of GPFS.So I think
the "tested" limit in 4.2 is 64 cores from memory (or was at the time of
release), but works just fine on 80 cores as far as I can tell.
JAB.
--
Jonathan A. Buzzard Tel: +44141-5483420
HPC System Administrator, ARCHIE-WeSt.
University of Strathclyde, John Anderson Building, Glasgow. G4 0NG
_______________________________________________
gpfsug-discuss mailing list
gpfsug-discuss at spectrumscale.org
http://gpfsug.org/mailman/listinfo/gpfsug-discuss
More information about the gpfsug-discuss
mailing list