Question How much faster should the RTX 4090 Ti be, over the RTX 4090?

Given the full-fat AD102 chip features 18432 shading units, 576 texture mapping units, 192 ROPs, 576 tensor cores and 144 raytracing acceleration cores, if history is any guide we can probably expect that or nearly so for any RTX 4090 Ti.

The gimped RTX 4090 is cut down to 16384 shading units, 512 texture mapping units, 176 ROPs, 512 tensor cores and 128 raytracing acceleration cores. This is considerably more trimmed down than GA102 was for RTX 3090, but that was kind of an odd case as 3080 and 3080Ti were also based on GA102 and presumably the ones trimmed a lot more to improve yields.

That alone is ~12% however the big question is the final clockspeed, which will certainly be set to whatever is required (and whatever wattage that means) to beat RX 7900 XT and any expected response in XTX or 7950 variants
 
We can assume a lot of things, from a lot of things, but we don't "know" anything, at all, until we do. And we don't, until the product reviews start popping up. Even what Nvidia (Or AMD) themselves "say" we can expect is often not exactly what the reality is when reviews are done.

Even after reviews come out it isn't always necessary exactly the final word on what a card can do or not do, because often their initial performance isn't fully accurate with what happens after various OS and driver changes occur.