Anyone have infomration on Star-CCM+ using Xeon PHI co-processor?

JoshuaB · November 15, 2013, 09:34

Hi Folks,

I was wondering if anyone has any experience/information about Star-CCM+ using a Xeon-PHI co-processor. I am in the midst of specifying a new computer and I saw this is now available. I have seen literature that indicates the present versions of Star-CCM+ do not see that great an increase with using a CUDA based GPU co-processor, but I was wondering if the results are any different with a XEON-PHI co-processor.

They (INTEL) advertises that one does not need to you recompile software to use it. But on the other hand the amount of RAM that comes with it seems limited and thus it might not be that much of a help when doing CFD types of problems (it might be constantly transferring information to and from its on-board memory to the computers main memory).

Any thoughts?

Thanks,

Joshua

TMG · November 16, 2013, 11:42

Both the Nivida GPU and the Intel PHI are different versions of the same animal. PHI is a little more flexible and easier to program but ultimately you still have to program for it.

STAR-CCM+ does not support either processor today because neither of them offer any benefits for the type of computing it does.

JoshuaB · November 18, 2013, 09:19

TMG, thanks for the reply, you helped to clear that issue up for me.

lordvon · February 4, 2014, 12:40

I do not think GPU and Xeon Phi are 'different versions of same animal'.

To use Xeon Phi, you only need to change some compiler flags, while for GPU you must completely rewrite the code.

Case in point, the guy in the following post used Xeon Phi for OpenFOAM by just recompilation (Of course optimization could be done by rewriting appropriate parts of code). Check it out:
http://www.cfd-online.com/Forums/ope...-xeon-phi.html

November 15, 2013, 09:34	Anyone have infomration on Star-CCM+ using Xeon PHI co-processor?	#1
JoshuaB New Member Joshua Brickel Join Date: Nov 2013 Posts: 26 Rep Power: 12	Hi Folks, I was wondering if anyone has any experience/information about Star-CCM+ using a Xeon-PHI co-processor. I am in the midst of specifying a new computer and I saw this is now available. I have seen literature that indicates the present versions of Star-CCM+ do not see that great an increase with using a CUDA based GPU co-processor, but I was wondering if the results are any different with a XEON-PHI co-processor. They (INTEL) advertises that one does not need to you recompile software to use it. But on the other hand the amount of RAM that comes with it seems limited and thus it might not be that much of a help when doing CFD types of problems (it might be constantly transferring information to and from its on-board memory to the computers main memory). Any thoughts? Thanks, Joshua

November 16, 2013, 11:42		#2
TMG Member Join Date: Mar 2009 Posts: 44 Rep Power: 17	Both the Nivida GPU and the Intel PHI are different versions of the same animal. PHI is a little more flexible and easier to program but ultimately you still have to program for it. STAR-CCM+ does not support either processor today because neither of them offer any benefits for the type of computing it does. JoshuaB likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
simpleFoam parallel	AndrewMortimer	OpenFOAM Running, Solving & CFD	12	August 7, 2015 18:45
DecomposePar unequal number of shared faces	maka	OpenFOAM Pre-Processing	6	August 12, 2010 09:01
[Other] StarToFoam error	Kart	OpenFOAM Meshing & Mesh Conversion	1	February 4, 2010 04:38
Unaligned accesses on IA64	andre	OpenFOAM	5	June 23, 2008 10:37
[Commercial meshers] Trimmed cell and embedded refinement mesh conversion issues	michele	OpenFOAM Meshing & Mesh Conversion	2	July 15, 2005 04:15

November 18, 2013, 09:19		#3
JoshuaB New Member Joshua Brickel Join Date: Nov 2013 Posts: 26 Rep Power: 12	TMG, thanks for the reply, you helped to clear that issue up for me.

February 4, 2014, 12:40		#4
lordvon Senior Member Robert Join Date: Sep 2010 Posts: 158 Rep Power: 15	I do not think GPU and Xeon Phi are 'different versions of same animal'. To use Xeon Phi, you only need to change some compiler flags, while for GPU you must completely rewrite the code. Case in point, the guy in the following post used Xeon Phi for OpenFOAM by just recompilation (Of course optimization could be done by rewriting appropriate parts of code). Check it out: http://www.cfd-online.com/Forums/ope...-xeon-phi.html