Imagine you could make your memory 1,000x faster, but your CPU, the memory controller, the numerous pathways and circuits between it and the CPU, and everything else in the system stays the same, you would be lucky to see an overall improvement in system performance of 25%.
This reminds me of various tests that people have done with performance versus HDDs, SSDs, and RAM Disks and load times (which is the biggest factor on why you want faster memory). The overall consensus that I found was even though the RAM disk showed something like 40-800 times the performance
over an HDD, loading times didn't decrease at that rate. What gives?
While I can't provide a definitive answer, one has to remember that Windows probably treats the RAM disk as any other disk. So on top of the file system operations that it must do (it makes software implementation easier if you unify everything), there some other things that people forget about. Loading isn't just putting the contents into RAM, it's also initializing data for program use. For instance, I'm fairly certain that loading an OS is not only memory intensive, but it's CPU intensive as well. In fact, Windows 8 shaved off a lot of time with the fast start up because it skips a lot of the initialization routines (their state gets stored).
To put this in another perspective, I work on microcontrollers and the initialization routines on them take about 91.5K cycles. And this may be a hard limit. So no matter how much or how fast the memory is, I have 91.5K cycles of boot time because the CPU can't go through it fast enough.