1600CL7 13560/9360
1873CL8 15100/10300 (adj. 13326/9090)
2000CL9 13480/9720
The cpuNB for the 1873CL8 tests is running 2273 MHz, not 2006 (stock) as in the other tests. So the cpuNB is ~13.5% faster and the scores average ~11.5% higher. You'll also note the Everest scores for 1600 and 2000 have a variance of ~2%±, which is not really statistically significant, and that the SiSandra scores are virtually the same at < 1%±. If you include the adjusted scores Everest results are ~3%± and SiSandra is, again, virtually the same at < 1%±. You might also note that while the other tests were run at 4.0+ GHz, the 1600CL7 tests were done at a CPU speed of 3.2 GHz, though the 3.6 GHz Turbo mode
may have kicked in, we don't know.
So what, exactly, did all that extra RAM speed (+400 MHz = +25%) actually do for performance?
Don't you find it interesting that the only real boost to RAM performance came from increasing the cpuNB speed?