Scrypt doesn't use any "RAM" in the sense you are thinking about. The Scrypt paremters chosen for LTC (and copied to every copycoin) use about 128KB of cache. Yes that isn't a typo kilobytes not megabytes or gigabytes. The only "memory" being used is the on die SIMD registers and low latency L2 cache. The two algorithm use the same resources, Scrypt simply requires more per thread and thus runs with a fraction of the parallelization that is possible with SHA-256. That combined with the increased number of operations in one hash explain the difference in throughput.
Then how does one get higher scrypt hash rate by increasing mem clocks, and why does GPU-z clearly show the allocation of 512 MB when cgminer starts?
Sure you could design an OpenCL kernel that mined Scrypt and SHA-256 based coins at the same time however you would be lucky to get 50% performance from each.
Perhaps, at a particular combination of core and mem clocks. The point is that to maximize scrypt hashing speed, one is forced to underclock the core down to 50 - 60% of mem.