CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > FloEFD, FloWorks & FloTHERM

Does Intel Phi helps on FloEFD?

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   May 27, 2016, 13:19
Default Does Intel Phi helps on FloEFD?
  #1
Member
 
Join Date: Nov 2014
Posts: 92
Rep Power: 11
hokhay is on a distinguished road
Hi all,

My company is looking for solution to speed up FloEFD simulation. Does anyone has an idea of Intel Phi works on FloEFD?
Besides adding more CPU, what else will benefit the performance?

Thanks a lot
Jason
hokhay is offline   Reply With Quote

Old   August 18, 2016, 04:45
Default
  #2
Disabled
 
Join Date: Jul 2009
Posts: 616
Rep Power: 23
Boris_M will become famous soon enough
Hi Jason,

No, Intel Phi is similar to a GPU card such as Nvidia Tesla family. The system of a processor in a PCI-E port is not supported.

In general the CPU generations increase in performance every time and the number of cores only help to some extent.
At some point the speed increase will level, for example using 40 cores would no speed up the solver much more than 32. It would make more sense to use a second solver for another project you need to solve and do it in parallel rather than in series.

So if you always want to achieve the fastest CPU time with FloEFD I can recommend to discuss with your team and management to annually buy new hardware and simply use the old hardware for other colleagues that need less resources. Basically handing them down every year.
Usually in companies there is a system upgrade of desktops every 3-4 years and you could try to turn it in the way that you get the best system every year a new processor generation comes out and give the old system to a CAD engineer who needs less performance for example and doesn't need always the newest system.

Also keep updating on the latest FloEFD version as new enhancements will potentially speed up the solver.

And lastly, try to create efficient meshes where it is necessary and try to save cells where they are not needed. For example I have seen models being setup with the automatic mesh settings and no local mesh which caused 3 times the mesh count than if the mesh was setup smartly to create an CPU time efficient mesh. Often the automatic settings can create a high small feature refinement at edges of a body where it is not necessary. In a 2D cut plot you can see the refining cells at the edge and often it doesn't look very bad but consider that the mesh is not 2D but this refinement of maybe 5 more cells in the 2D plot stretches along the whole edge of the model and cause easily 200,000 cells more just on the one edge.

Mentor is constantly trying to improve the solver speed and works with Intel to better leverage the Intel CPU technology. V15 has increased in solver speed and newer version will continue to push the limit.

I hope this helps,
Boris
Boris_M is offline   Reply With Quote

Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
[OpenFOAM.org] OpenFOAM build on Intel Xeon Phi asaijo OpenFOAM Installation 31 July 26, 2017 11:35
[OpenFOAM.org] Trouble Compiling OpenFOAM-dev using Intel Compiler 15 for use on Xeon Phi foamer123 OpenFOAM Installation 9 August 20, 2015 15:03
openFoam, Intel Phi, without intelcompiler/intelmpi m0guib OpenFOAM Installation 2 November 3, 2014 08:05
[OpenFOAM] Color display problem to view OpenFOAM results. Sargam05 ParaView 16 May 11, 2013 01:10
CFX11 + Fortran compiler ? Mohan CFX 20 March 30, 2011 19:56


All times are GMT -4. The time now is 05:35.