CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > OpenFOAM > OpenFOAM Running, Solving & CFD

Openfoam parallel calculation performance study - Half performance on mpirun

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

Like Tree3Likes

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   March 23, 2018, 19:39
Default
  #21
Retired Super Moderator
 
Bruno Santos
Join Date: Mar 2009
Location: Lisbon, Portugal
Posts: 10,969
Blog Entries: 45
Rep Power: 127
wyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to all
Quick notes:
  1. Read this thread:OpenFOAM benchmarks on various hardware
  2. The bottleneck here is the number of memory channels, not the number of cores. Even if your CPU had 64 cores, it will not perform better than 8 cores, if there are only 4 memory channels available.
lourencosm likes this.
__________________
wyldckat is offline   Reply With Quote

Old   March 24, 2018, 04:28
Default
  #22
New Member
 
Jurado
Join Date: Nov 2017
Posts: 22
Rep Power: 7
Jurado is on a distinguished road
Quote:
Originally Posted by wyldckat View Post
Quick notes:
  1. Read this thread:OpenFOAM benchmarks on various hardware
  2. The bottleneck here is the number of memory channels, not the number of cores. Even if your CPU had 64 cores, it will not perform better than 8 cores, if there are only 4 memory channels available.
Hello wyldckat,

Thanks for your answer, your thread is very instructive. But I am not sure to understand what are these "memory channel", I tried to look on google to try to understand it but I don't really do, could you explain it to me ? It is hardware right ? Is it the thing on which one connect the RAM ? ( Sorry i'm really not an expert in Hardware).

Also how can I know how much memory channel I have ? Do I have to open the computer ? ( Well i guess with your explanation i must have 10 used)



Than you.
Jurado is offline   Reply With Quote

Old   March 24, 2018, 20:40
Default
  #23
Retired Super Moderator
 
Bruno Santos
Join Date: Mar 2009
Location: Lisbon, Portugal
Posts: 10,969
Blog Entries: 45
Rep Power: 127
wyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to allwyldckat is a name known to all
Quick'ish answer:
From a previous post, you stated that the following CPU model is being used: Intel(R) Xeon(R) CPU E7-4870 @2.40 GHz
You can find the specific details for it at Intel's website, specifically on this page: https://ark.intel.com/products/53579...-GTs-Intel-QPI
Quoting from that page:
Quote:
Max # of Memory Channels: 4

[Description]
The number of memory channels refers to the bandwidth operation for real world application.
Wikipedia article on the topic of multiple memory channels: https://en.wikipedia.org/wiki/Multi-...y_architecture

Google for "memory channels explained" to find more details. First video on the topic: https://www.youtube.com/watch?v=-D8fhsXqq4o


But to explain in a somewhat simple way, by way of comparison:
  1. If you yourself have 2 arms, it's as if you had 1 CPU (your own brain) and can pick up 2 packages at the same time (one package with each arm).
  2. Now if you are surrounded by 4 other people around you, that ask you to pick up 1 package for each one of them, this means that you will need to pick up 4 packages... but you only have 2 arms, so you will have to do at least 2 trips to bring them the packages they asked.
  3. Now, even if you had 4 arms, and have 9 people around you, and if each one asks you to bring 1 package and if you need another package for yourself, this means that you need to pick up 10 packages, while only using 4 arms, means you need to do 3 trips to go get the packages.
Now, while you physically could use some additional hardware to make the job easier, the CPUs cannot do that. The E7-4870 CPU only has 4 memory channels and that means that 10 cores are sharing those 4 channels.


Still confused? Here is another thread with more details: AMD Ryzen Threadripper vs Intel Xeon, importance of cache and memory channels
nimasam and lourencosm like this.
wyldckat is offline   Reply With Quote

Reply

Tags
openmpi, parallel calculation, performance

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
Map of the OpenFOAM Forum - Understanding where to post your questions! wyldckat OpenFOAM 10 September 2, 2021 05:29
How to contribute to the community of OpenFOAM users and to the OpenFOAM technology wyldckat OpenFOAM 17 November 10, 2017 15:54
OpenFoam parallel on 2 computers : Cannot find file "points" Blue8655 OpenFOAM Running, Solving & CFD 1 June 3, 2015 21:59
Superlinear speedup in OpenFOAM 13 msrinath80 OpenFOAM Running, Solving & CFD 18 March 3, 2015 05:36
Performance of interFoam running in parallel hsieh OpenFOAM Running, Solving & CFD 8 September 14, 2006 09:15


All times are GMT -4. The time now is 00:30.