CPU utilization during optimization

Igor Zakharov:

Most likely, the tester does not re-count the passes that have already been made. They have already been passed. And regarding the memory/CPU balance-it’s the same
optimization, there is a probability (high!) the fact that the passes are uneven in load, i.e. at some point again there may be a percentage of 5%
memory 90, and maybe Vice versa )) Although, such a number of cores should average everything. But maybe it does, is it a genetic test or
full…

But this is all speculation, only you know for sure what the robot is doing. )))

As I understood in the end: at the initial start of optimization, the load on the cores is unevenly distributed (on Xeon e5 and DDR3 I try for the first time and before
that – on 4-cores and DDR4 this was not the case). Most likely, at the start, there is not enough operatives for everyone (but is it so and why then everything is evenly distributed?) the
Result is a simple core and they ignore the total load on the “brothers” – no more than 10 (+/- ) remain in the work. They don’t reconnect themselves.
Uniform loading occurs after stopping and restarting. What is interesting is that the percentage has a frequency of 2.4 when all cores are running. Now everything
works, but the frequency is 2.8 when loading at 82 percent (which is strange) and the tester gave the operation-70 percent against 90 on the first runs.