This forums's habit of stripping the context of quotes strikes again; smokeown was not really asking if the 2GB units will be slower than the 4GB units; he was asking whether hypotherical 2GB units with half the memory bandwidth would be slower than 4GB ones with full bandwidth.
And the answer is; probably, yes. Maybe 10% of instructions read or write RAM in my experience, and as I understand it, this memory bandwidth would mainly be gained when reading or writing sequential blocks, such as the framebuffer. Things like the stress test that doesn't output anything for seconds at a time, and other working memory would mainly fit in the cache though probably wouldn't be as effected, but anything that uses the screen (like a game) or uses large data-sets would almost certainly be significantly slowed by only having half of the sequential bandwidth. Block copy using normal 32-bit ARM instructions would run at most at 32-bits per two clock cycles, which I assume would be faster than the maximum memory bandwith in dual channel mode, so it's simply down to how fast the memory runs how fast that will complete.
