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

Relationship about mesh number and CPU working performance

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

Reply
 
LinkBack Thread Tools Display Modes
Old   November 1, 2012, 21:53
Default Relationship about mesh number and CPU working performance
  #1
New Member
 
Fatih
Join Date: Apr 2012
Posts: 4
Rep Power: 5
geforce is on a distinguished road
Hi

I have two different case about mesh number. In One of them, I have 1.5 million mesh number, In other, have 13 million mesh. My CPU is Intel 3930k which has 3.2 ghz with 12 core and Memory is 16 gb.

When I run the calculation which has 1.5 million mesh, My CPU always runs %100 and my calculation is more speed than other calculation naturaly. But When I eun other calculation which has 13 million mesh, My Cpu sometimes run %100. Especially, When It starts to new iteration, It waits so much. Computer has been using any CPU for two or three minutes and not always use %100. CPU usage become so changeable. But for 1.5 milion mesh, It always works %100. What is the reason or reasons of this situation.

Sorry my weak english.
Regards,
geforce is offline   Reply With Quote

Old   November 2, 2012, 04:10
Default
  #2
New Member
 
Karl Kargl
Join Date: Mar 2009
Location: Austria
Posts: 9
Rep Power: 8
Karl is on a distinguished road
Your PC does not have enough memory for the large case. This is the problem - it uses the virtual memory during the run. Therefore you can see the unbalanced load in the system monitor. I think you need at least 24 or 32 GB of memory (depends on the physical models of your case).

Best Regards,
Karl
Karl is offline   Reply With Quote

Old   November 2, 2012, 06:22
Default
  #3
New Member
 
Fatih
Join Date: Apr 2012
Posts: 4
Rep Power: 5
geforce is on a distinguished road
Thank you for answer KARL.

I use K-epsilon turbulance model and Non-equilibrium wall function. What is the relationship between mesh number and ram size. For example, I can say that maximum 7 milion elements can be solved with 16 gb ram. Do turbulance models affect the maximum mesh number. virtual memory means, Computer use the different components for the memory, like hard-disk. So, data speed of the between ram and processor is slow and processor waits the mesh informaiton for the next process.
geforce is offline   Reply With Quote

Old   November 2, 2012, 15:07
Default
  #4
Member
 
vicarious's Avatar
 
Pedram Mojtabavi
Join Date: Apr 2011
Location: Iran
Posts: 66
Rep Power: 6
vicarious is on a distinguished road
Send a message via Yahoo to vicarious
The more mesh cells the more RAM size you need. When the number of the cells are too much and your RAM has not enough memory, it takes more time the data to get loaded in the RAM for every iteration, therefore your CPU has a delay to get the data. Turbulence modeling get more CPU time.

Best regards,
vicarious is offline   Reply With Quote

Old   November 2, 2012, 16:33
Default
  #5
New Member
 
Fatih
Join Date: Apr 2012
Posts: 4
Rep Power: 5
geforce is on a distinguished road
Quote:
Originally Posted by vicarious View Post
The more mesh cells the more RAM size you need. When the number of the cells are too much and your RAM has not enough memory, it takes more time the data to get loaded in the RAM for every iteration, therefore your CPU has a delay to get the data. Turbulence modeling get more CPU time.

Best regards,
Thank you vicarious, I want to ask something. My mesh files 900 MB, but Ram usage is almost 16 GB ram. Why is main reason of this.
geforce is offline   Reply With Quote

Old   November 2, 2012, 17:15
Default
  #6
Member
 
vicarious's Avatar
 
Pedram Mojtabavi
Join Date: Apr 2011
Location: Iran
Posts: 66
Rep Power: 6
vicarious is on a distinguished road
Send a message via Yahoo to vicarious
Your mesh size is not equal to the RAM's memory that has been occupied. Every iteration produces a large amount of data that get loaded as a very big matrix into RAM. It contains the values of every parameters in 2d or 3d and their first, second or higher order derivatives, so it gets more space. For 13 mil grid, a larger RAM would have less delay.

Regards.
vicarious is offline   Reply With Quote

Old   November 5, 2012, 08:08
Default
  #7
New Member
 
Fatih
Join Date: Apr 2012
Posts: 4
Rep Power: 5
geforce is on a distinguished road
Quote:
Originally Posted by vicarious View Post
Your mesh size is not equal to the RAM's memory that has been occupied. Every iteration produces a large amount of data that get loaded as a very big matrix into RAM. It contains the values of every parameters in 2d or 3d and their first, second or higher order derivatives, so it gets more space. For 13 mil grid, a larger RAM would have less delay.

Regards.
Thank you so much for your reply.
geforce is offline   Reply With Quote

Old   November 5, 2012, 09:10
Default
  #8
Member
 
vicarious's Avatar
 
Pedram Mojtabavi
Join Date: Apr 2011
Location: Iran
Posts: 66
Rep Power: 6
vicarious is on a distinguished road
Send a message via Yahoo to vicarious
Quote:
Originally Posted by geforce View Post
Thank you so much for your reply.
You're quite welcome.
vicarious is offline   Reply With Quote

Reply

Thread Tools
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 On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Superlinear speedup in OpenFOAM 13 msrinath80 OpenFOAM Running, Solving & CFD 18 March 3, 2015 06:36
Large Eddy Simulation with too high Mach number Roland R CFX 4 November 25, 2010 17:43
air bubble is disappear increasing time using vof xujjun CFX 9 June 9, 2009 07:59
Structured mesh refinement Andrea Panizza FLUENT 1 November 9, 2003 04:48


All times are GMT -4. The time now is 17:15.