Most of the numbers on the spec sheet are useless for comparison. Things like memory bandwidth, texture and pixel rate are calculated from specs (bit depth times memory clock or cores times clock speed). That means nothing in the real world when architecture and memory algorithms affect real world performance. The number of cores isn't relevant like a cpu where something uses a certain number of threads because a gpu is much more paralleled. You can't just look at clock rates, memory speed or bit depth when in the end, they're only parts of total bandwidth/performance so all of them have to compared in total.
Specs can't be compared on different architectures which makes comparing them pointless when a higher model tends to be better...