I know there are alot of discussions already on performance, and no one answer is going to solve every problem. However - I have not been able to find a discussion that seems to fit my scenario.
Problem statement:
Linear Array feature with multiple intersecting cuts.
Rebuild time can take upwards of 2 minutes in some cases.
Array is 20x20 (400 instances)
CPU usage of 13%-15% when rebuilding a part - and no, there is not one CPU topped out at 100%. Graphics card frequency will drop to 50MHz (from 475 MHz) when rebuild is 75% complete.
Computer specs:
Custom built.
Gigabyte X58A-UD7
i7 950 - socket 1366 - OC to hit 4.1GHz (water cooled)
12 GB DDR3 @ 688 MHz
Quadro 4000
Questions:
Should I expect better performance?
Are there tips for improving performance - other than remove hyperthreading and checking the use of OpenGL?
Your linear array is only using one core. Yes, it will float between the different cores, but it is still just using one. It will switch between using each one to balance out the thermal load.
Should you expect better performance? I don't know, but creating arrays off things is one of the most processor intensive things you can do. So this is probably true. Try other methods of modeling if you can. Can you upload an example of the kind of thing you are trying to array? Perhaps there is a modeling technique to solve your problem.
I'm scared about your statement "checking the use of OpenGL". What does that mean? Do you have "use software OpenGL" checkmarked? I hope not. This is really a misnomer. That checkbox should be called "disable the video card and only use software OpenGL".