If you read through the thread, 0.325v is exactly the safety differential between vcore and 1/2 vdimm. And yes, without re-reading all the thread, IIRC someone arrived to that conclusion the same way as you, CnQ vcore versus JEDEC 2.85v gives you 0.325v.

With all this said, what do you meant with "that logics does not work"? You are clearly stating that it works with your 2 next statements

So AMD safety boundis that, 0.325v with max JEDEC specs