Maximising Output Question

[Dave442]

Hi all,

I have been given a HP workstation which has a quad-core Xeon 3GHz CPU, 16GBs of ram and a 64bit windows architecture to carry out some research using ABAQUS/Explicit v6.9. I currently have access to 10 Abaqus/Explicit tokens and am trying to determine how to maximise my output on this platform.

As far as I can figure out I have the option of queing jobs to run two at a time (2 jobs = 2 x 5 tokens = 10 tokens) using a single core each or to run one at a time in parallel (1 job = 5 + 3 x 1 token per extra cpu = 8 tokens) across all four cores.

Using a very small test model I can run 2 seperate jobs in about 1.25 times the time it takes to run a single job in parallel. Im not sure however how the simulation time for running a job in parallel scales with increasing degrees of freedom.

If anybody could provide some insight or advise it would be greatly appreciated!

Many thanks,
Dave

 

[rstupplebeen]

For a low number of cores I have almost always found it to be most beneficial to run with the maximum number of CPUs. The only exceptions I can think of are very small problems where PRE is a significant portion of the analysis or in your case where you are using all cores so there is none open for the OS, CAE or other programs. In the end the best tests are running your actual problems and see. You might want to try 3 CPUs also.

If your motherboard can support it I would suggest getting a second quad-core. When compared to the cost of Abaqus it should not be a difficult cost justification. I hope this helps.

Rob Stupplebeen

https://sites.google.com/site/robertkstupplebeen/