From a post I made on B3D; I do find the whole multiples of 18 system very elegant:
On PS4 (2013):
18CU @ 800MHz - PS4 | Base Mode
On PS4 Pro (2016):
18CU @ 800MHz - PS4 | Base Mode
18CU @ 911MHz - PS4 | Boost Mode
36CU @ 911MHz - PS4 Pro Patched
On PS5 (2020):
18CU @ 800MHz - PS4 | Base Mode
18CU @ 911/2233MHz - PS4 | Boost Mode
36CU @ 911MHz - PS4 Pro Patched
36CU @ 2233MHz - PS4 Pro Patched | Boost Mode
36CU @ 2233MHz - PS5 | Base Mode
And thinking ahead, for eg.
On PS5 Pro (2024):
18CU @ 800MHz - PS4 | Base Mode
18CU @ 911/2233MHz - PS4 | Boost Mode
36CU @ 911MHz - PS4 Pro Patched
36CU @ 2233/2500MHz - PS4 Pro Patched | Boost Mode
36CU @ 2233MHz - PS5 | Base Mode
36CU @ 2500MHz - PS5 | Boost Mode
54CU @ 2500mHz - PS5 Pro Patched
On PS6 (2028):
18CU @ 800MHz - PS4 | Base Mode
18CU @ 911/2233MHz - PS4 | Boost Mode
36CU @ 911MHz - PS4 Pro Patched
36CU @ 2233/2500MHz - PS4 Pro Patched | Boost Mode
36CU @ 2233MHz - PS5 | Base Mode
36CU @ 2500/3300MHz - PS5 | Boost Mode
54CU @ 2500mHz - PS5 Pro Patched
54CU @ 2500/3300mHz - PS5 Pro Patched | Boost Mode
108CU @ 3300MHz - PS6 | Base Mode
On PS6 Pro (2032):
18CU @ 800MHz - PS4 | Base Mode
18CU @ 911/2233MHz - PS4 | Boost Mode
36CU @ 911MHz - PS4 Pro Patched
36CU @ 2233/2500MHz - PS4 Pro Patched | Boost Mode
36CU @ 2233MHz - PS5 | Base Mode
36CU @ 2500/3300MHz - PS5 | Boost Mode
54CU @ 2500mHz - PS5 Pro Patched
54CU @ 2500/3300MHz - PS5 Pro Patched | Boost Mode
108CU @ 3300MHz - PS6 | Base Mode
108CU @ 3600MHz - PS6 | Boost Mode
162CU @ 3600MHz - PS6 Pro Patched Mode
That said, by the time PS7 comes around and assuming it's a traditional, local piece of higher spec hardware. I wouldn't be surprised if we see some paradigm shifts and considerable changes in architecture. Eventually maintaining hardware blocks or cells on the die that support multiple past systems may begin to prove an outsized detriment to native software, as well as having a bloated, complex array of modes to maintain. The good thing is from PS4 onwards games are x86 based and from PS5 onwards they're unlikely to have logic tightly coupled to exact clocks. So the prospect of shifting some elements of BC into the software emulation (along with the hardware power to brute force it) may not be so daunting as it is for current, past systems.
As for PS1/PS2/PS3/PSP/PSVita. I'd love to see them circumvent the entire issue of BC and just build a miniscule version of the CELL B.E., PS1/PS2 logic vital for perfect execution of that software and a basic ARM chip; all on a tiny die made on modern processes. If we see future platforms move to chiplets this could be included as an extra chiplet. The issue of course would be having adaptive timings/latencies/clocks given that circuits would be much, much shorter on such a small die. I think I've said this before, but they could just get in at the base level and address the problem there -- sooner rather than later -- rather than expend vast resources on complex, imperfect software solutions that will require per-software checks and fixes. This chip could be sold standalone in a budget legacy console, as a marketable extra in Pro consoles (as an add-on chiplet on SoCs); and embedded in future Portables/Handhelds. This "PS123-PV" die could theoretically probably be <10% the size of a modern APU die in consoles.