All these shameful L3 caches are exactly the same crutch as in cheap SSDs, where they write huge numbers of read/write speeds only within the SLC cache. As soon as the SLC cache runs out, cheap SSDs immediately drop to a level that is often worse than SATA3 max read/write speed. The situation is approximately the same with all x86, which was completely disgraced against the background of Apple's M2 Max memory controller, which is 6 times faster than AMD memory controllers and 5 times faster than Intel.
As soon as this "3DCache" runs out (and today it is not even enough to keep all the key system drivers inside, not to mention software and games), any x86 processor immediately becomes about the same memory speed as a SATA3 SSD compared to 990 Pro and older.
x86 is at a complete standstill, and for several years now the kernels have been suffocating from the monstrously slow RAM.
Why do new processors get pci-e 5.0 28+ lines, if the top AMD processor can only serve 12-14 at a time?
When will we all consumers see RAM expandable in slots with read/write speeds of at least 300Gbyte/s?