Error while running the 1M element neon model

[gkaf89]

I am trying to run the HPC 1M element neon benchmark model, and the solution terminates early with the error message:

MESSAGE ID :        205  

 ** RUN KILLED: ENERGY ERROR LIMIT REACHED


MESSAGE ID :        207  

 ** RUN KILLED: TOTAL MASS ERROR LIMIT REACHED

 NC=  124700 T= 6.0001E-02 DT= 1.6117E-09 ERR= 99.9% DM/M= 5.0701E+29
 ELAPSED TIME=        524.74 s  REMAINING TIME=        174.90 s
     ANIMATION FILE: NEON1M11A014 WRITTEN

The output file from the starter and engine executable is the following: SLURM_NEON1M11_868775.out.txt

The issue was detected in the following releases:

• latest-20230719
• latest-20230720

The execution scripts used are the following:

• batch_script.txt
• openradioss.txt

The bug appears in the following 2 clusters:

1. Intel cluster (Skylake CPUs):

• 4 nodes
• 8 MPI processes
• 14 threads per process

2. AMD cluster:

• 4 nodes
• 32 MPI processes
• 16 threads per process

The attached output is from the AMD cluster.

[roguic]

Hi @gkaf89 ,

Unfortunately the Neon model is a sensitive model. According to the hardware/openradioss version, you need to play with the /DT/NODA/CST parameter in the *1.rad.

You can try :
/DT/NODA/CST/0
0.6 5e-007

or
/DT/NODA/CST/0
0.67 6e-007

If you want to test the scalabity/performance of OpenRadioss, maybe the Taurus 10M element model is more convenient (https://openradioss.atlassian.net/wiki/download/attachments/47546369/Taurus10M.7z?api=v2).

Best,
Romain

[gkaf89]

Hi @roguic,

Thank you very much! I will close the issue given that this behavior is known.

Best,
Georgios