Everest Ultimate Edition 2006:
The memory benchmarks in Everest Ultimate Edition can provide four sets of results: memory read speed (in MB/s), memory write speed (in MB/s), memory copy speed (in MB/s), and memory latency (in ns). Higher values are better for memory read, write, and copy speeds, while lower values are desirable for latency. Here we see that even though it holds a performance edge in all tests and configurations, the DDR3 performance doesn't start to shine until you get to 1375MHz or above. Considering the rather vanilla nature of this DDR2 kit, it is easy to picture a higher speed, lower latency kit giving the Kingston PC3-11000 a run for its money.
Sisoftware Sandra Lite 2008.1.12.34:
The memory benchmarks in SiSoftware's Sandra were used to analyze three more sets of results: integer (in MB/s), float (in MB/s), and latency (in ns). Higher values are better for integer and float, while lower values are better for latency. Again we see that it is close when the DDR2 kit is at its highest speed and the DDR3 is at its lowest speed, but that the gap widens significantly when the DDR3 is at or above its rated speed.
Performance Test v6.1:
This final suite of tests is new to me, and I decided to run them after seeing them used on a few other sites. Performance Test v6.1's memory benchmarks provide six different values, and higher is better for all of them. Unlike the first two sets of tests, here we see that the DDR3 kit doesn't hold an advantage over DDR2 800MHz or DDR2 984MHz until it is overclocked to 1400MHz. While I do not have much history with this test, it is interesting to see another perspective on the performance of the kits tested.