Unless your Mac OS X 10.6 build is set to use the 64-bit kernel already, when upgrading to Mac OS X 10.7 you will find huge performance gains in CPU intensive tasks like code compilation. When building ImageMagick with GCC 4.2.1 from Xcode (and this test profile using the number of make jobs set to twice the number of available CPU threads), for example, the timesavings was huge. The 32-bit 10.6.0 SMP kernel was slower than the single-threaded 11.0.0 kernel that is using the 64-bit build by default. Once Apple releases the finished Mac OS X 10.7 with a working 64-bit SMP kernel, the performance improvements will be even greater.
In a single-threaded benchmark like BYTE, the Dhrystone 2 performance was much greater, of course, thanks to the default x86_64 kernel.
When Mac OS X 10.7 is officially released, we will also run benchmarks of Mac OS X 10.6 from the 32-bit default kernel configuration and then again, when manually switching to the 64-bit kernel so that there can be a direct 64-bit 10.6 vs. 10.7 comparison. However, even right now with the architecture and threading differences, the results are looking very favorable for Mac OS X 10.7. C-Ray is one of our favorite multi-threaded ray-tracing benchmarks and even right now with the current Lion limitations, its numbers are looking quite favorable.