LS-Dyna runs giving different results on new run every time with same settings

@anuragdogra

Simply use diff utility in VIM or any other modern editor and ensure the job files are indeed one and the same. You can do the same thing via LS-PrePost (Model Compare).

Thanks @icebreakersours! But my models look exactly the same between the one I submitted to the cluster and the one I have on my local machine that was working fine. Is there anything else that could lead to such behavior?

Best

Anurag

All I meant to suggest you want to double-check your belief that the models are exactly the same. I know I have held that belief many times in the past only to realize that that was not the case.

That being said, while rare, it is possible that some features can give diverging results depending on the hardware. I don't remember it off the top of my head but an openly available projectile simulation has a contact card that gives dramatically different results between Windows and Linux. If you switch the card to a more hardware-agnostic card, the results look the same.

So, that said, take a step back and ensure there are no variables (like OS, CPUs, MPI, etc.) and repeat runs with the exact same keyword deck on the same hardware. Then, change one variable at a time (say, more cores). Proceed accordingly.

Hi @anuragdogra, 

If the files are the exact same between the local and cluster/HPC simulation, then it might be something to do with the cluster and not your model. What error exactly are you getting when you try to run on your cluster?