[Pierre1]

When I use Workbench to run my CFX simulations, I check the task manager. Then I see, not only CFX solver programs occupy the power of the CPUs, but also the Workbench (figure attached).

This means, for example, if I have 12 cores to run 3 Workbench simulations, I’d better set the MPI processors number to be 3, instead of 4, to all of them because each of Workbench program (in task manager it shows as ‘AnsysFWW.exe’) use at least one additional core.

I’d like to confirm here that why this thought is correct?

[ghorrocks]

The AnsysFWW.exe task should not take significant CPU time or memory during a simulation. You do not need to allocate it a CPU.

But having said that, this is why I run CFX stand-alone. So you don't have all the overhead of workbench hanging around.

[Pierre1]

Originally Posted by ghorrocks

The AnsysFWW.exe task should not take significant CPU time or memory during a simulation. You do not need to allocate it a CPU.

But having said that, this is why I run CFX stand-alone. So you don't have all the overhead of workbench hanging around.

But you will have to use 'Workbench' if you have a lot of design points to run.


For me, I always need to run hundreds of design points. The time consuming will kill me if I change parameter settings in cfx pre for each of them.

[ghorrocks]

You can run zillions of runs in workbench. You can also use shell/batch files or batch processing software. There are many ways of running lots of runs. And yes, shell/batch files and batch processing software have ways of handling variations between the runs as well.

Have a look at the CPU and memory load of the AnsysFWW.exe task during a test simulation. Then see for yourself whether it needs to be allocated resources or not.