Is there supposed to be some surprise at the fact that 16GB outperforms 8GB in synthetic benchmarks? Your "test" defines nothing really. Both are important. More memory, to a point, will always result in higher scores on synthetic benchmarks, but if you tested two 8GB sets of modules with the same frequency (2666mhz) and one set had CL16 while the other had CL14, the CL14 sticks will always show some improvement over the higher latency sticks. It's pretty common knowledge. There is no mystery here.