Hi, for 3d explicit dynamics (autodyn), I have 48 virtual processors available. How to use them all at full potential to reduce the solution time. Thanks a lot.
For 16K elements, please try 2 cores.
HPC performance is highly model-dependent. Running time is usually reduced if more CPUs are used. However, if model size is not large, running time may not get reduced if a high number of CPUs are used. This is because more CPU time is spent on exchanging data between CPUs, rather than model solving at each CPU. Usually the best HPC performance can be reached if each CPU has the number of elements 10,000~100,000 to solve.
For each HPC job, Autodyn solver spawns a master job plus the specific number "n" of slave jobs. The total number of running jobs in your computer is "n+1". If you have 48 processors and use all of them for HPC, the total number of the running jobs is 49. This means that there are 2 jobs are competing each other for the same CPU. This will deteriorate HPC performance.
The recommended practice is to use 46 or less CPUs. So the total number of the running jobs is 47 or less. This will not jam your computer and thus obtain the best performance.
Thanks @cxquan , for this helpful answer.
My problem size is small approx 24k nodes and 16k elements, but the time step is very small approx 2e-8 and I need at least 300 of such steps to fully capture the simulation.
And it's demanding a tremendous amount of time, although I'm using 46 v.cores, it's asking for 200+ hours, is it normal for a material cutting simulation on a mm scale ?
Dammnn @cxquan ... Magic happened!!! 200+ hrs to 72 hrs approx. Thanks a lot!!!
Ansys customers with active commercial software licenses can access the
customer portal and submit support questions. You will need your active account number to register.