Yeah, I'm not even sure that anything I do is bandwidth limited on dual-channel 1333MHz DDR3. Quad-channel 2011 platforms seem to get similar scores if you look at similar core counts & clockspeeds as if more-than-doubling the memory bandwidth is irrelevant
They don't in the memory bandwidth benchmark!
Small benchmark programs are likely to fit well in large caches and with HT there to allow useful work to be done, when a core is blocked waiting for memory access mitigating load/save stalls. The story on access times, may be different particularly on not very multi-threadable poorly tuned software which may not have good locality of reference, have less predictable access reducing benefit of out of order, or have cache line bouncing.. for example perhaps because things like garbage collection are going on.
For mainstream systems, that people buy, they're probably over-whelmingly limitted most of the time by access to HDD storage, or in certain programs by RAM capacity. The idea of "good enough" applies. If you've really got specialised requirements, then evaluating by specific benchmark is the best answer. Gamer benchmark stuff is showing the graphics card ought have much more of the budget than the CPU, if you're maximising value and "balancing" the system.
I think most people intuitively over rate the importance of CPU speed, though when specc-ing a new system the marginal cost of faster is often justifiable, most CPU upgrades are probably vanity driven, rather than cost effective especially compared to storage upgrades like SSD's.