KZ: SF does ray casting, on screen space, for local reflections. That's way less demanding than a ray tracing renderer, and that other engines like Cry Engine 3 also does (even on last gen consoles, if I'm not mistaken).
But even if it wasn't the case, pooling the framebuffer for more data does not makes the framebuffer larger, it would make the total bandwidth usage increase, but the buffer size should be the same. (that is assuming that the performance issue xbone is seeing is because people are simply not being able to fill all their buffers into the esram)...
My main point was that right now there's clearly a performance issue, that's undeniable, but I don't think there should be, not to that degree at least... But like I said before, Ms has created for themselves some major potential performance blackholes: All the hardware access is virtualized, on top of that they have a VM running on top of another VM, the gpu has to serve two different OSes at the same time, there's that talk that xbone gpu drivers are a lot closer to W8 than 360, and has huge overhead... It's hard to point the culprit without knowing exactly what is causing the issue, but looking at the numbers, I really don't think it should be struggling this much (Specially because it is struggling with some last gen ports too)