Multithreading: Prallelization, Setup to balance / calculate optimal settings and more


this might be more a general question. There is a setting about multiple threads for all nodes.

When executing a rather compute intensive task, I notices my 16 Core 32 Thread AMD Ryzen 7950X barely moved the pinky.

Using five threads via Parllel Start / End Nodes improved CPU utilization as expected.

The question is, how does the setting about “Maximum working threads for all nodes” get’s determined and how does it work / scale with the parallel start / end nodes?

Reason for my question is that I feel there is quite often performance left on the table. I can imagine, as the question about Knime tuning quite often is asked, to create standardized performance evaluation workflow which, after it was executed, suggests some optimal settings.