I am not sure we are speaking about the same thing 
The way 3D cache is distributed across will indeed have performance implications in some use cases.
For reference here is the output of lstopo for 5950x, showing how cache is distributed:
Here you can see that 5950x is a dual CCX architecture with 16 threads in each. Unlike 5800X3D, which is a single CCD.
7800X3D is also a single CCD.
According to AMD it will come with 104 MB of total cache (L2+L3), which works out to 1 MB L2 cache per core and 96 MB of L3 cache.
In their announcement, dual-CCX 7900X3D and 7950X3D were shown to have caches of 140 MB and 144 MB, while they would have been 204 MB or 208 MB if it was distributed evenly across all cores like in 7800X3D.
Therefore in 7900X3D and 7950X3D the additional 3D Vcache will be attached to one CCX, not the other. It is an asymmetric cache setup. Different CCXs do not share L3.
That’s why 7800X3D is likely the sweet spot for MSFS, and some other games as well…
For workstation usage, typically, it is another story.
I for one am planning to use virtualization extensively, so in this case the higher end CPUs can be better suited as well