I've yet to see any differnece in memory benchmarks with different timings. With large amounts of RAM, (1024 and up), slower timings are usually necessary for stability anyway, and with larger amounts, slowing the timing might reduce latency and improve performance. Looking at the post you linked, It's hard to get very excited about the benchmarks. I just use the fastest stable timings. Running the FSB too fast can also decrease memory performance, contrary to popular opinion.