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

not all memory is used

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

Reply
 
LinkBack Thread Tools Display Modes
Old   March 17, 2011, 22:48
Unhappy not all memory is used
  #1
New Member
 
Join Date: Mar 2011
Posts: 13
Rep Power: 6
Romek is on a distinguished road
Hi,

I am trying to run a simulation with Ansys 13 on a Server with 16gb memory.

The simulation requires about 12gb, but i cant run is. It says not enough free memory allocated. Then I reduced the solver memory allocation step by step to 0.7 where it needs about 8.5gb memory, but same problem appears.
I can run a smaller simulation with 7gb memory allocated.

Why cant I use the whole 16gb (minus what the OS needs)?
Is it an Ansys settings problem?
Romek is offline   Reply With Quote

Old   March 22, 2011, 20:53
Default
  #2
New Member
 
Join Date: Mar 2011
Posts: 13
Rep Power: 6
Romek is on a distinguished road
please does someone have a hint?

i still didn't figure out how to solve this problem. The system is running on Windows Server 2008 R2.

i can run a simulation with 2 fluid domains and 2 fluids specified (needs ~8gb). When i try to add a 3rd domain and a 3rd fluid it's not working for the reason i mentioned. I have enough memory available.

if you need more information just tell me
Romek is offline   Reply With Quote

Old   March 22, 2011, 22:28
Default
  #3
Super Moderator
 
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 10,818
Rep Power: 85
ghorrocks has a spectacular aura aboutghorrocks has a spectacular aura aboutghorrocks has a spectacular aura about
Are you running CFX or ANSYS FEA?

In CFX you can always just do a parallel simulation and bring the memory problems under control that way.
ghorrocks is online now   Reply With Quote

Old   March 22, 2011, 23:04
Default
  #4
New Member
 
Join Date: Mar 2011
Posts: 13
Rep Power: 6
Romek is on a distinguished road
I'm running CFX in HP MPI local parallel mode with double precision.

I tried to go down with the solver memory allocation factor as far as possible but no luck...
Romek is offline   Reply With Quote

Old   March 23, 2011, 09:13
Default
  #5
Senior Member
 
Join Date: Apr 2009
Posts: 513
Rep Power: 12
stumpy is on a distinguished road
It not enough memory has been allocated then you should increase the memory allocation factor, not decrease it. If not enough memory was available then that's a different problem. How many partitions are you using? If it's allocating more than 4GB per partition for any single stack then it could be due to a bug - contact support for a fix.
stumpy is offline   Reply With Quote

Old   March 24, 2011, 00:02
Default
  #6
New Member
 
Join Date: Mar 2011
Posts: 13
Rep Power: 6
Romek is on a distinguished road
thank you so much, i solved the problem with your help!

I didn't have a partition problem before. The partitions were indeed larger than 4gb because I was using 2 partitions.

Increasing the number of partitions finally made the solver run...

I can use 3 or 4 partitions (more has got me license issues), which number is better? I'd say 3 is better due to less larger stacks, right?
Romek is offline   Reply With Quote

Old   March 24, 2011, 14:28
Default
  #7
Senior Member
 
Join Date: Mar 2009
Location: Europe
Posts: 168
Rep Power: 8
joey2007 is on a distinguished road
Unless you are above 100k-200k nodes per partition you can use as much partitions as possible. Below this number of nodes per partition it will get inefficient.

On quad core machines the band width to memory may be a bottle neck. Recommendation: on many machines it will be siginificantly more efficient to use only three of four cores.
__________________
-
-
-
-
-
------------------------------------------------------------------------
Please do not forget: I am not paid for answering your questions.


Thousands of issues can cause a division by zero. Please do not capture a thread, with the argument: "I have the same issue ...."
joey2007 is offline   Reply With Quote

Old   March 24, 2011, 20:36
Default
  #8
New Member
 
Join Date: Mar 2011
Posts: 13
Rep Power: 6
Romek is on a distinguished road
I have got 1.5m nodes and 3 partitions is about 6% faster per iteration step than 4 partitions. Just tested on 1st time step!
Romek is offline   Reply With Quote

Old   March 25, 2011, 14:46
Default
  #9
Senior Member
 
Join Date: Mar 2009
Location: Europe
Posts: 168
Rep Power: 8
joey2007 is on a distinguished road
Roughly spoken: the solution process is parallized. Anything before and after is done by the master process alone. In order to measure the speedup you have to compare wall clock time of the solution process. Otherwise the i/o will bias the result.
__________________
-
-
-
-
-
------------------------------------------------------------------------
Please do not forget: I am not paid for answering your questions.


Thousands of issues can cause a division by zero. Please do not capture a thread, with the argument: "I have the same issue ...."
joey2007 is offline   Reply With Quote

Old   July 7, 2011, 19:38
Default
  #10
New Member
 
Mridul
Join Date: Mar 2011
Location: Melbourne, Australia
Posts: 26
Rep Power: 6
makkks is on a distinguished road
Hi Guys,

I m using Ansys 13 SP2, 256GB ram, 48 core .
the simulation is giving memory allocation error.
sum Errors are not even on CFD site like:

Stopped in routine MEMPAR
Stopped in routine ILOCCS

I tried to change d factor in partitioner, solver but nothign is helping.

sometimes it is saying
"not enough memory was available"

what should I do?

I m using MPCH2 LOcal Parallel with 44core for this simulation.
seems each core is picking 8GB
reducing slover allocation factor is not working,

Plz help
makkks is offline   Reply With Quote

Old   July 8, 2011, 06:58
Default
  #11
Super Moderator
 
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 10,818
Rep Power: 85
ghorrocks has a spectacular aura aboutghorrocks has a spectacular aura aboutghorrocks has a spectacular aura about
You have some serious hardware available there. Have you tried doing test runs with smaller numbers of processes?
ghorrocks is online now   Reply With Quote

Old   July 10, 2011, 18:21
Default
  #12
New Member
 
Mridul
Join Date: Mar 2011
Location: Melbourne, Australia
Posts: 26
Rep Power: 6
makkks is on a distinguished road
Thanks for the reply Glen,

yup I have tried to use it with small processes and it works fine.

Also, I just tried running again and changed the Partitioner to 0.8x , Solver to 0.4x, Interpolator to 0.4x and finally it started running but after 2 steps it stopped and gave error again saying

Stopped in routine MEMERR.

It used upto 120GB memory and 48 cores in the above process.
makkks is offline   Reply With Quote

Old   July 10, 2011, 19:04
Default
  #13
Super Moderator
 
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 10,818
Rep Power: 85
ghorrocks has a spectacular aura aboutghorrocks has a spectacular aura aboutghorrocks has a spectacular aura about
I do not have access to that many processors or licenses. You are seeing problems I cannot possibly see, so I do not think I can help you. As you must have a significant investment in this setup I would contact CFX support for help to get it running ASAP.
ghorrocks is online now   Reply With Quote

Old   July 10, 2011, 19:07
Default
  #14
New Member
 
Mridul
Join Date: Mar 2011
Location: Melbourne, Australia
Posts: 26
Rep Power: 6
makkks is on a distinguished road
Thanks a lot for your support Glen
makkks 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
Fluent Parallel Memory Usage David Christopher FLUENT 1 March 3, 2011 03:20
How to optimize the memory usage when using FEM vasilis Main CFD Forum 11 August 24, 2009 23:57
Memory error? Young CFX 3 September 30, 2008 11:33
CFX CPU time & real time Nick Strantzias CFX 8 July 23, 2006 17:50
Run-Time memory conifguration Error Graeme CFX 1 February 21, 2006 00:04


All times are GMT -4. The time now is 02:36.