CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > General Forums > Hardware

Workstation for post-processing of large cases

Register Blogs Community New Posts Updated Threads Search

Like Tree2Likes
  • 1 Post By flotus1
  • 1 Post By flotus1

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   August 29, 2018, 12:11
Default Workstation for post-processing of large cases
  #1
New Member
 
Join Date: Dec 2017
Posts: 10
Rep Power: 8
RedoFromStart is on a distinguished road
Hello!
I am in need of a workstation capable of post-processing large cases - usually between 70 and 100 million hexahedral cells.
The software I am going to use is Fluent and CFD-Post.
My budget is about 8000 Euro.
Please ask if any further information is needed.

Last edited by RedoFromStart; August 29, 2018 at 16:37.
RedoFromStart is offline   Reply With Quote

Old   August 30, 2018, 04:40
Default
  #2
Super Moderator
 
flotus1's Avatar
 
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,399
Rep Power: 46
flotus1 has a spectacular aura aboutflotus1 has a spectacular aura about
You are probably looking for these specs:
  • A single CPU with only one NUMA node, decent clock speeds and around 8-12 cores. E.g. Xeon W-2145.
  • At least 128GB of ECC memory. 4x32GB and a motherboard with 8 DIMM slots if you stick with a Xeon W.
  • 2TB or more of NVMe-based storage
  • A good Nvidia graphics card with plenty of VRAM, e.g. Quadro P4000 or P5000.
  • Depending on how you get the data to this workstation: 10GBit Ethernet or more.
Gerhard likes this.
flotus1 is offline   Reply With Quote

Old   August 31, 2018, 14:27
Default
  #3
New Member
 
Join Date: Dec 2017
Posts: 10
Rep Power: 8
RedoFromStart is on a distinguished road
Can you suggest any other CPU?
The one you specified is hard to get in my country.
RedoFromStart is offline   Reply With Quote

Old   September 3, 2018, 06:25
Default
  #4
Super Moderator
 
flotus1's Avatar
 
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,399
Rep Power: 46
flotus1 has a spectacular aura aboutflotus1 has a spectacular aura about
Hmm...you could drop down to consumer CPUs like the I7-7820x or I9-7900x.
But these do not support more than 128GB of RAM (8*16GB) and also lack ECC support.

I don't feel qualified to judge whether one of those might be a deal-breaker for you. I honestly don't know how much RAM CFD Post will use for your applications.

Reading the initial post again got me thinking: Is this thread for a "general purpose" workstation that will also run simulations or is it specifically for post-processing? My initial recommendation was for a dedicated post-processing workstation. It severely lacks performance for CFD simulations and does not have enough RAM to run most simulations with this cell count. E.g. Fluent coupled solver in double precision uses 3-4GB of RAM per million cells.
flotus1 is offline   Reply With Quote

Old   September 3, 2018, 17:42
Default
  #5
New Member
 
Join Date: Dec 2017
Posts: 10
Rep Power: 8
RedoFromStart is on a distinguished road
The workstation is specifically for post-processing.
Simulations are performed on a cluster.

I think that CFD-Post has similar requirements for hardware as Paraview if it helps you with judgement.

Is ECC crucial for such task? From what I've read its usually slower than normal RAM.
RedoFromStart is offline   Reply With Quote

Old   September 3, 2018, 18:11
Default
  #6
Super Moderator
 
flotus1's Avatar
 
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,399
Rep Power: 46
flotus1 has a spectacular aura aboutflotus1 has a spectacular aura about
The performance difference for ECC vs. non-ECC is in the low single-digit percentages for the same rated speed.
What is different is that with an I7 or I9 CPU you could use faster memory. Xeons these days are completely locked, so DDR4-2666 is the highest possible memory speed. Other current-gen CPUs can go way beyond that.

Do you need ECC or not... controversial question. Some folks would burn you at the stake for even mentioning non-ECC memory. Others are running entire clusters with non-ECC memory without any issues.
In my opinion it depends on the scope of application. If you are in a larger company, consult your IT department. They might have a policy concerning this topic. Otherwise ask yourself the worst-case question: how bad would it be if this workstation did an unintentional crash, losing a few hours of work. Or if it produced a bogus result due to a flipped bit. How much money is involved in the results you produce? Are the results relevant for safety of people or machinery?
If you can take those risks in exchange for a cheaper and/or faster machine, non-ECC might be for you.
Gerhard likes this.

Last edited by flotus1; September 4, 2018 at 06:14.
flotus1 is offline   Reply With Quote

Old   September 4, 2018, 08:08
Default
  #7
Senior Member
 
Joern Beilke
Join Date: Mar 2009
Location: Dresden
Posts: 498
Rep Power: 20
JBeilke is on a distinguished road
It very much depends on what type of data you have (steady or transient) and what type of postprocessing you want to do:


- calculating numbers
- section and vector plots
- volume rendering
- Virtual Reality


512 GB of main memory is probably not too much :-)
JBeilke is offline   Reply With Quote

Reply


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
Parallel processing of OpenFOAM cases on multicore processor??? g.akbari OpenFOAM Running, Solving & CFD 31 November 1, 2017 09:25
Dual cpu workstation VS 2 node cluster single cpu workstation Verdi Hardware 18 September 2, 2013 03:09
Pre- and Post workstation - need inputs larsenmm Hardware 17 February 13, 2013 04:16
Post Processing of Turbine blade cases screech1987 Main CFD Forum 0 March 31, 2011 18:59
Post Processing in FEM Abhijit Tilak Main CFD Forum 0 April 26, 2004 11:59


All times are GMT -4. The time now is 07:44.