I'm running JMP 10 and am having some issues with performance. Monitoring system performance during creating a Space-filling Latin Hypercube DOE shows that CPU usage is at 25% and approximately 1 GB of RAM remains free. Is there a way to encourage multicore threading and high memory usage to speed up such calculations?
In addition, it seems there is an issue with memory allocation, as my RAM usage remains high even after such a process has finished.
I'm running with 4GB of RAM with an Intel Core i5 @ 2.50GHz processor.
I've noticed that JMP doesn't like to free up memory after getting ahold of it. I specifically was looking at memory usage associated with opening large tables and closing them, the additional memory used to have the table open was not all released to the system when the table was closed. I also noticed that JMP 10 uses about 8% more memory to have files open than JMP 9.
Not an answer to your question, but a confirmation of some of the issues you're experiencing.
One question - do you have 4 cores? I ask because I also see my CPU frequently pegged at 25% overall when JMP is crunching, but it is all concentrated on a single core, which is maxed out at 100%. Now, I can pretty much do anything else outside of JMP simultaneously since I have 3 cores left to work with, but that's not what I need.
It seems that JMP does not support multi-core operations, even some that would obviously be improved (i.e. column formula calculations when not dependent on row, running matrix operations).
I would like to see JMP multi-core support, and how to enable it in JSL. Otherwise, my new AMD 4 core desktop is reduced to a fast 2005-era single core chip
I have a dual-core processor and I never see JMP use more than the one core. SAS is barely multi-thread enabled (in fact, I believe most major statistical packages are not enabled - I know R can implement it, but isn't automated) so I wouldn't hold my breath on JMP getting it enabled even though it would be great to get multi-threads on things like All Subsets.
A few operations in JMP are multi-threaded, though it's mostly higher end analytics. Most sorting is multi-threaded, but you needs tons of rows to notice it. Distribution is multi-threaded if there are multiple columns. Try Distribution on 8 variables gets my 8-core machine briefly to 100%.