So as I understand it Sampler Feedback is available on all DX12U hardware and it's just the "streaming" part that's unique to the XSX. And the streaming part equates to custom hardware to blend the transition between the current mip and the new mip if the new mip doesn't arrive in time for when it's needed.
So if my understanding is correct then it begs the question: how is basic Sampler Feedback supposed to work without that hardware to blend the mips? Was it implemented in hardware with the expectation that we'd simply have to live with sudden jarring transitions? Or was there an expectation that as you say, the blend could be done in software via shaders?
Well I'm out on a branch here speculating, but from reading the Sampler Feedback stuff for DX12U I think on PC you can get, literally feedback on what was hit in memory and what was missed. So even with just that, through either shaders or the CPU with API feedback you could then choose what to do, and then implement some kind of software driven mitigation strategy. It's unlikely that Nvidia DX12U stuff - as awesome as it is - has an equivalent to MS's custom hardware solution for seemingly "free" handling of transitions. Obviously, the more that's handled in software, the greater the overheads are likely to be. But with a bit of time, a bit more hardware, and some creative thinking the PC has proven for about the last 25 years that it'll find a way in some form or another...
Now on XSX, I *think* that through some combination of hardware and API back end you can set the system up to automatically request and then begin to blend in the higher-res mip map once it's no longer "missed". We know from MS that hardware is definitely involved in managing both the fetching and the blending, so there reasonably has to be more to it than just Sampler Feedback and Trilinear filtering. Maybe it's a hardware shortcut to skip writing a shader or CPU code to handle it, freeing up the overhead and reducing latency. I dunno!
So in other words, in DX12U on PC, I'm
guessing that the Sampler Feedback data on individual page misses or hits is also available, but that more is left up to the developer to manage / evaluate through either shaders or the some kind of feedback to the CPU. Remember that this isn't all on MS through the DX12U API, it's also on Nvidia, Intel and AMD, their respective hardware, and what they want to implement in their APIs. And some of the stuff that can support this (well, basically Nvidia) is knocking on for a couple of years old now and a lot older than MS and their SFS reveal.
So in other words, if you know what developers on console are likely to want to do with Sampler Feedback on your particular console, you can probably build in some specific hardware acceleration. But there's normally another option, especially if time is on your side ...