CPU OC or not for MSFS 2020?

Here's CPU performance of my 10700k running MSFS 2020 flying from KBUR to KSAN, which is a flight path nearly right over download Los Angeles, CA.  The results of three flights are given below, all were made using the same few clouds static weather setting and time of day, and using the Windows 10 Balanced power plan. The three plots were snapped close within vicinity of downtown LA.  The MSFS 2020 graphic settings were nearly full Ultra, other than both LODs at 150, and Clouds and Textures at High.  See my specs in my sig, using 30 fps external lock.  The first run was made using completely stock bios settings.  The second's only bios changes were an all core 47 ratio (a 4.7Ghz OC) with MCE explicitly disabled, and the third is my full up 5.0Ghz OC.

The most surprising difference in these three runs is that full stock settings seems to show a better use of the cores.  But looking into it closely, this seems more of an artifact of the Windows 10 scheduler getting excited about its "favored core list" whose processors' desirability may be further enhanced by Intel's "turbo boost max technology 3.0".  This "excitement" is causing threads to be bounced in and out of the first four logical cores on a regular basis, hence its jagged saw tooth appearance.  This behavior is consistent from run to run, though the four logical cores used can vary, e.g 4-7 instead of 0-3.  Beyond what I stated is completely conjecture, does this now really seem that much better of a cpu core usage scenario?

The second two plots potentially show a grim reality for MSFS in the form of a big fat main thread.  This behavior is completely consistent across multiple runs.  I suspect that when the bios core ratios are not set to auto, "turbo boost max technology 3.0" is disabled which in turn effectively disables the Windows 10 scheduler's excited use of the "favored core list", thus the threads are no longer being tossed from core to core.  It a bit comforting to see how my OC has tamed the main thread somewhat, but dang, even so there is NOT that much headroom there, and that's running only at 30 fps on nearly the fastest CPU available with 4000 cas 16 ram and a 3.4/3.1 GB/s read/write M.2 drive for the program and cache.

Dang, I REALLY hope better MSFS 2020 runtime optimizations are in the works.

Any thoughts, or better interpretations of what these plot have to say?

Thanks for reading.

Edited October 5, 20205 yr by TheFamilyMan