[dvy]

I am facing some problems with memory allocation at the HPC cluster at my institute.  Each node in this cluster has available a memory of 44gb and eight CPUs (NProcShared=8). So if I increase the number of nodes (LindaWorkers) the amount of memory available increases by 44 gb for each additional node.

Now, if I try to assign on a single node for an input :
 %mem=44GB

I get the error:
Number of steps in this run=    648 maximum allowed number of steps=    648.
 Out-of-memory error in routine Optmz4 (IEnd=    6770633057 MxCore=    5905580032)
 Use %mem=6457MW to provide the minimum amount of memory required to complete this step.
 Error termination via Lnk1e in /opt/software/g09/l103.exe at Wed Mar  6 04:08:32 2013.
 Job cpu time:  0 days  0 hours  6 minutes 17.0 seconds.


When I try to use 2 nodes (I should have 88 gb memory available to me now) with the input line :
%mem=44GB

I get the same error:
Number of steps in this run=    648 maximum allowed number of steps=    648.
 Out-of-memory error in routine Optmz4 (IEnd=    6770633057 MxCore=    5905580032)
 Use %mem=6457MW to provide the minimum amount of memory required to complete this step.
 Error termination via Lnk1e in /opt/software/g09/l103.exe at Thu Mar  7 10:42:59 2013.
 Job cpu time:  0 days  0 hours  6 minutes  3.6 seconds.


When I try to use 3 nodes (I should have 132 gb memory available to me now) and change the input line to:
%mem=132GB

This ends with error:
Leave Link    1 at Fri Mar  8 01:49:58 2013, MaxMem=17716740096 cpu:       0.8
galloc:  could not allocate memory.

Also:
Operating System: Red Hat Enterprise Linux Server release 5.4
Processor :Intel Xenon Nehlam Processors (2.93GHz) with 8 cores

Can anyone help me out with this?