Maybe the memory overclocking is something that only works if and the cpu and the motherboard, support it.
That plus i really have a problem with the umbrella term "memory overclocking". There are two things that can be called "memory overclocking":
1. Making memory run at the speed greater that stated at the memory specification (JEDEC spec or POR)
2. Making memory controller operate at the speed greater than specified in the CPU specification (because memory controller is part of the CPU now)
For the latest - the situation with the fourth gen. Asus WS and AsRock WS boards is quite unique - i do not remember last time any motherboard allowed memory controller on a Xeon Scalable processors (i mean server chips) operate at anything beyond what is advertised in the CPU specification - because usually the maximum frequency of the memory controller is locked in the CPU level. And i worked with gen 1, 2 and 3 processors before - ES, QS and retail versions alike.
For the first - making memory run at the speed greater than this memory is specified to - i did it many times with no problems, i ran DDR4 2133 at 2666 with first generation of Xeon Scalable, i ran DDR4 2400 at 2933 with second generation and i'm running DDR4 2666 at 3200 with both Xeon 3rd gen and AMD EPYC 7742 right now - and i haven't encountered any motherboard - server or workstation - who does not allow ignore POR specification and run memory at the maximum speed supported by CPU. May be some HP or Dell workstation motherboards have that option locked down but who cares about those.
So i think we just got lucky that for some reason some processors (Xeon Scalable gen.4 D0 stepping) have unlocked memory controller and some motherboards (like Asus WS) allow exploit it to get speeds far beyond specified in the ark.intel.com for that CPU model. That is just a happy coincidence and i'm grateful for it but treat it as such - happy glitch that works in our (consumers) favor.