scrypt-N is a misguided attempt at being ASIC-resistant.
It's quite possible to make ASICs that handle any scrypt-N
with N in the range (10..20), since one can feasibly fit 128MB
on a single chip.

At N=20 however, scrypt is already unusable for another reason,
namely that verification effort also scales exponentially with N
(a fact conveniently ignored by its proponents).

X11 is something of a joke. 11 hash functions that are each
designed to have efficient ASIC implementations, are still
very ASIC friendly in combination.

The claimed GPU benefits are due solely to suboptimal implementation.