Link: Max Cores
At the risk of repeating the question above, now I’m being more specific:
- How do I diagnose the reason that fewer than 100% of the cores are being used?
The model is:
- Sequential lens
- Merit function is the default minimum spot size with constraints on thickness / radii
- There are (x3) configurations for thermal
I am seeking Global Optimization solutions.
I have a “new” computer:
- (x32) Cores
- 128 GB RAM
When I started the optimization last night, the fan started spinning so I knew (?) it was working hard. This morning, all is quiet. (x7) of (x32) Cores are working on the problem and I have very little physical RAM in use as reported by the Windows Resource Monitor.
Referencing the question above:
- The number of assigned variables during the optimization (x20 variables and fewer than x20 cores used)
- OpticStudio will only use as many cores as the number of variables assigned
- The amount of RAM in your system (Way more memory available than cores)
- For each variable assigned, OpticStudio loads a copy of the entire system into memory before tracing/ evaluating.
- Systems with imported CAD parts can commonly run out of RAM before utilizing all available cores.
- To check the estimated memory required per copy of your system:
-
- For each variable assigned, OpticStudio loads a copy of the entire system into memory before tracing/ evaluating.
- Last, some Analyses/ Merit Function Operands can only be executed single-threaded (All Analyses windows are closed except the Spot Diagram and Layout. Which ones are single-threaded?)
- This could result in intermittent reduction in core usage during an optimization cycle.
- Closing Analysis windows did not increase the number of cores used
- Turning off Auto Update, increases the number of cores used for Hammer Optimization
- Turning off Auto Update for Global Optimization still doesn’t increase the number of cores to (x20 = number of variables)
- Just started another Global Optimization. (x19) variables. Right at the start, all (x32) cores are being used
- Does this imply that the number of cores is related to the progress in searching? or the remaining size of the optimization space that has been explored?
Best answer by Kevin Scales
View original