ANSYS R16 slower than R15 by 30% in nonlinear contact analysis.

[assafwei]

Hi,

I started using R16 in the last few days.
I was doing a non linear contact problem (with nonlinear materials and displacements).
I used a pure penalty contact type and on gauss point detection method.
I ran the exact same analysis on R15 and on R16 and got 30% slower solve time on R16.
Could this be some kind of withdrawal in the ansys solver performance or am I doing something wrong?
BTW - I am using an HPC pack in both setups (same number of cores used in both setups).

Thanks.

 

[Reitzel]

Hi there

Are you positive that you have applied the same Solve Process Settings in v. 16 as in v. 15, i.e. number of utilized cores, GPU accelerator etc? Did you solve the model on the same computer?

 

[assafwei]

Hi,

All parameters are tha same on both setups.
Solver type is program controlled in both, I manually selected contact penalty and detection method, including manual stiffness and penetraion tolerance.

 

[Reitzel]

Hi

Alrihgt. Just be aware that, if you just installed v. 16, the Advanced Solve Proces Settings (accessed by clicking Advanced under Solve Process Settings) have been reset. So if you, or someone else, has entered Advanced Solve Proces Settings and clicked the GPU accelerator option (requires license to run) or increased the maximum number of utilized processors from the default 2 to e.g. 4 (might also require a license, not sure though) then that could explain the difference.. Otherwise no clue!

 

[assafwei]

Hi,

Both in R15 and R16 the Advanced Solve Proces Settings are set the same (8 cores).

 

[pthieffry]

This is indeed surprising, as the opposite would be expected with R16 - feel free to join our webinar about R16 on Wednesday, there is a portion on contact that will show advances in performance. See

https://marketing.ansys.com/rc/ansysinc/tp/15070

For your current probleme, I would recommend you to contact your local support and share your model with them.