EPYC based workstation for OpenFOAM simulations

flotus1 · May 30, 2018, 08:49

Mixing ranks is definitely the worst idea. If you want to use different RAM for whatever reason, at least make sure the DIMM organization is identical.
Btw, I don't see how my last comment concerning RAM contradicts what I wrote previously.

klilla · May 30, 2018, 10:29

Dear All, who try/use epyc processors with openfoam. What is your experiences? I mean, I have tested two epyc workstations and, well, they behave strangely with openfoam.

I ran an LES test case with interFoam - so VoF - with 28M cells. I tested intels and amds, different ages, I know it is not fair, but I tested everything I have/had access to. Now, all of them is scaling, except epyc. I could push this test case with close to linear scaling up to 512 processors, meaning, that epyc should really scale beyond 32 - which is the 'CPU increase'=1 in the attached figure.

Do you have other experiences?
Thanks,
Lilla

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Strange residuals of the Density Based Solver	Pat84	FLUENT	0	October 22, 2012 16:59
amd opteron based workstation needed	ztdep	Hardware	1	December 5, 2010 07:23
incorrect temperature in pressure based solution	Kian	FLUENT	1	July 6, 2009 06:59
Absorption coeff:Cell or domain based for boiler	DEV	FLUENT	1	October 1, 2008 07:57
Pressure based and Density based Solver	Xobile	Siemens	1	November 30, 2004 22:13

May 30, 2018, 08:49		#21
flotus1 Super Moderator Alex Join Date: Jun 2012 Location: Germany Posts: 3,427 Rep Power: 49	Mixing ranks is definitely the worst idea. If you want to use different RAM for whatever reason, at least make sure the DIMM organization is identical. Btw, I don't see how my last comment concerning RAM contradicts what I wrote previously.