Surely if you can plot temps over time, you can do same with the CPU GHZ, or? (Mine is locked to 1.6 Ghz, and temp is constant running MIP1 only, no different to what it shows for OET1 and ZIKA)

Hi SekeRob,
unfortunately, I do not plot the CPU frequency. However, the plot of the used memory is available, see below:

If the CPU has to wait regularly on the memory (for example because of recurrent cache faults), the CPU load would remain at 100%, even if the CPU is waiting (waiting is not equal idle), but, because waiting is less demanding than crunching, the CPU temperature will drop down.
I don't know how the science software is designed and implemented, nevertheless, my feeling is that something could probably be optimised on software level.
Comparing the CPU temperature with the memory load, my guess is that there is a possible cause for the crazy MIP behaviour on Linux.
If not MIP, it is a library or a kernel problem on Linux level (Linux 4.4.0-93-generic #116-Ubuntu SMP Fri Aug 11 21:17:51 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux).
Cheers,
Yves

I am switched over a Linux system from running SCC to MIP and noticed no difference in temperatures. Psensor is giving me temps of 37C to 44C on a system which runs 14 to 16 hours per day.
Cheers

Understood. The part I can't reconcile is the temp of the core going down and utilization staying the same (everything else being equal such as fan speed, ambient temp etc). Was your utilization graph only for core 0 or was it total utilization across all cores? I could understand one core dropping in temp but would expect it's utilization to also drop. while the other cores were not impacted. In that case, maybe MIP is causing some sort of on chip cache pollution that only effects one core. Whereas OET and other apps don't do that. I believe there tools out there to analyze the cache but that is not something I would care to undertake. If you have multiple machines with different processor models, do those other machines suffer the same malady? Maybe it is only related to a particular processor model or family. I had an E5560 that would not run correctly after Ubuntu 15.04. With 16 FAH jobs running, some cores would run 60 percent, a few others would be 40 and the rest 100. E5550 was OK, E56xx series was also OK. Only the E5560. It wouldn't run on 15.04, 15.10, 16.04. Problem went away with 16.10. I suspect it was firmware for that model that was dropped in 15.04 and corrected in 16.10. I think I was able to tie it back to a specific ABI level of the 4.8 kernel.

In the second link I posted above on updating 16.04 LTS to 16.04.02 there's a command that also pushes hardware support improvements

sudo apt install --install-recommends xserver-xorg-hwe-16.04

Ran it because ever since 16.04 my screen would cycle, when GUI loaded, to power off / power on after the designated delay time to lock screen. Remember, my kernel has been stepped up over time to now 4.13. After running the command, the problem is gone, and to my surprise it installed another kernel, 4.10, i.e. a fully updated 16.04.02 uses 4.10, but after boot I chose for 4.13 again, and the screen problem remained gone.

edit: spell daemon