Does a 20% overclock usually yield a 20% performance increase?
depending on how the architecture and new hardware features go, we could see 1:1 or even greater than 1:1 scaling since 'instruction prefetching' is clock based and not time based. so similar to 'branch prediction' it often gives a bigger performance boost as clocks increase.
As in, being clock based means it always takes a certain number of cycle to perform an operation, so the higher the clock the more operations that can be performed in the same amount of time.