There’s no reason to assume frontier-level intelligence eventually collapses all the way onto a midrange consumer GPU. In fact, there are quite a few reasons not to assume that (information-theoretic constraints, etc).
But, it could happen for a coding-focused model, or an accounting-focused model, etc. most tasks only need a subset of the total model to be done effectively.
There's no information theoretic constraint we know of that prevents this. You will almost surely win a Turing award if you can prove this.
It's almost a given that whatever is frontier intelligence today will run on a potato in a few years.