Hi Adisa,

It is not a good idea to include all available cores in the solve process settings. You should leave a few for other processes for best performance. Also note that ANSYS does use physical cores not logical processors that are shown in the image, so in your case you should set the cores to 2 or max 3 on your machine for better performance.

Regards,

Aniket

Hi Adisa,

You have properly setup ANSYS to use all your available cores.  The solver consumes the core, which spends very little time waiting for data, so ANSYS is programmed to submit jobs to cores, not threads. That means you will only see about 50% on your performance meter that counts threads not cores.

I have a desktop computer that is dedicated to solving ANSYS models and I turned off Hyperthreading so there is only one thread per core and therefore my performance meter does reach 100%.  I have a laptop to do my other work.

To anchavan's point about leaving one core to do other things: Mechanical provides two options that can be configured differently. Notice the pull down menu next to Solve. It lists two options: My Computer and My Computer, Background.  You can configure the first with 4 cores and the second with 3 cores. When you decide you want the solution as soon as possible, and you will perhaps walk away from your computer or just wait, you click Solve and the default is My Computer. But if you want to use your computer to do other things while you wait for the solve to run, you can pull down and select My Computer, Background. This behaves a little differently to Solve, My Computer since when the Background Solve finishes, you have to right click on the Solution branch and Get Results.  However, you can close Mechanical and even Workbench while the solver is running in the background.  When you open Mechanical later, you can Get Results.

Hi Adisa,

Optimizing solution time can be complicated.

Have you looked at the Solution Output? Below is an example.

You can see the Elapsed time for this job was 568 seconds or 9.8 minutes.

Let's say you have a model than runs in 10 minutes using all 4 cores on your 4 core machine.

Let's say you have a variation on that model and you want to solve that also.

If you solve them sequentially, that will be 20 minutes of total elapsed time.

Try solving them simultaneously. Maybe it will come in below 20 minutes, but maybe it will come in above 20 minutes.

I guarantee that you won't get them both solved in 10 minutes. The only way you could do that is if you had 8 cores and assigned each job to use only 4 cores.

MEMORY CONSIDERATIONS

You can see that on this computer I have 16 GB of physical memory (RAM) available, but this job only needed 3.1 GB of RAM.

Pretend the job that ran in 10 minutes on 4 cores needed 14 GB of RAM. Even if I had 8 cores, but only 16 GB of RAM, running two jobs that each need 14 GB of RAM on a 16 GB machine is not going to finish in 10 minutes because the solver will run out of RAM.  The only way the two jobs are going to finish in 10 minutes on an 8-core machine is if the machine has 32 GB of RAM.

I have more discussion about memory and cores at this discussion.