From the PrimateLab benchmarks and relative to the 3.06 GHz Core 2 Duo, we would get the following.
Overall Performance:
Core I7 @ 2.66 GHz: 29% faster
Core I5 @ 2.53 GHz: 18% faster
Core I5 @ 2.40 GHz: 14% faster
Integer Performance:
Core I7 @ 2.66 GHz: 18% faster
Core I5 @ 2.53 GHz: 7% faster
Core I5 @ 2.40 GHz: 3% faster
FP Performance:
Core I7 @ 2.66 GHz: 34% faster
Core I5 @ 2.53 GHz: 23% faster
Core I5 @ 2.40 GHz: 17% faster
Keep in mind the limitations of any benchmarks. They only measure what they're designed to measure. For example, in evaluating the above figures, consider:
1. I don't believe that benchmark is multithreaded, so if your app or OS is multithreaded, there will be some additional benefit because of hyperthreading.
2. I assume that that benchmark would take advantage of TurboBoost. If it doesn't, then the actual results would be a little better.
3. That benchmark is mostly CPU performance. Keep in mind the differences in the other parts of the system. With OSX, for example, some things that were typically considered CPU dependent will benefit from the faster GPU in the new systems.
For those reasons, I rarely consider synthetic benchmarks (Geekbench, for example, shows single digit percentage increases on systems that are massively faster in real life). The BEST benchmark is to find the applications you use and compare them. Apple almost exclusively uses application benchmarks. They're not perfect, but IMHO, they're better than a synthetic CPU-only benchmark.