Indeed yes it seems like the dual channel 6000MHz wasn't the limiting factor in many* tests. But also the lower power 395 has a couple massive wins, and I suspect it was almost entirely the memory bandwidth. For example the LLM test on page 2: it was that memory bandwidth making that colossal difference.
There were also tests like OpenVINO where memory bandwidth was the factor, but the x3d's cache was evidently big enough, with it easily scoring first place, the Strix Halo second, then the regular 9950 last place!
In a lot of tests, it seems like the extra 100w the 9950's were getting probably helped a lot!
Scaling to PCIe 6 is more bandwidth yes. But HBF is supposed to be going far faster, is the perception.