That analysis is incredible, how was the OP able to figure all of that out!?

I've experienced some early adopter pains with the desktop version of this CPU (x299 platform) when it came out and I used linux kernel mailing list to search for answers. I was surprised how easy it was to read the code of the related kernel modules and interact with their authors (Intel engineers).

These "magic codes" are probably just #define macros somewhere in the kernel repo, in my case it was in [1]

[1] https://github.com/torvalds/linux/tree/master/drivers/platfo...

I don't know, but apparently (1):

>> Originally posted at MacRumors by winterny >> https://forums.macrumors.com/threads/optimal-cpu-tuning-sett...

1. https://news.ycombinator.com/item?id=17602455

The Reddit post was about 20 minutes earlier.

Lots and lots of testing and lots of reading intel docs / example code on the internet. I probably ran cinebench and prime95 about 2000 times in the past few days.

Intel's HDK docs?

If you're designing hardware, this is a thing that people know. The new Xps machines have the same VMR problem. Probably he tested for a known problem and bingo! I wonder if the discoverer works for any hardware design company.

If you read that thread, what's curious is that it seems this power threshold has always been set perilously because older MacBooks seem to have the same setting for their VRMs as well.

I would expect this sort of slip from Dell or Lenovo with like 1200 skus but Apple really only has a handful.