Very slow Newton iteration

[kvslab]

Very slow Newton iteration

[jorgensd]

Could you link me to the specific example you are running.
It is weird that the residual doesn't decrease with a recomputed jacobian.

This would be expected if and only if we use an iterative solver with a high tolerance to solve for the Newton step.

It could be worthwhile changing the relaxation factor of the Newton step as well.

[keiyamamo]

Hi @jorgensd
Thanks for the reply!
Here is the link to the example, but it is an aneurysm case that needs Saga to reproduce the problem. I’m not even sure if the problem (increase in the residual) is reproducible. It might work if I resubmit the job on Saga, so I will probably try that and perhaps ask more help to you if I confirm that the problem is reproducible.

I did not think about changing the relaxation factor but I’ll definitely try that. That way I may be able to avoid the problem. Thank you so much for sharing your insight!

Best,

[keiyamamo]

Haven’t experienced the same issue recently. Closing for now.