Finding the best ram timings for any given setup is a trial and error process. Lower the timings, test for stability. If it passes, do it again, if it fails, back it up a notch.
it is kind of a tiring process, OCing ram, as there are so many timings to fiddle with, and each time a single number is changed, stability should be tested. If stability fails after changing more than one variable, you cannot know which variable is to blame.
Also, the tighter (lower) your timings are, the less likely you will be able to lower them further, as each change is a greater % of the current. Lowering CAS9 to CAS8 is approximately an 11% drop in timings. Dropping CAS4 to CAS3 is a %25 difference. I would consider 4-4-4-12 timings to be rather tight as is. Higher voltage can sometimes (not always) stabilize tight timings, but then you start running into danger scenarios with overvolting memory.
Than being said, the difference between tight ram timings, and absolutely tight as possible ram timings, is not noticeable to anyone but extreme benchmarkers. If you goal is to see actual differences in general performance, you probably won't. As such, I would assume if you were into hardcore benchmarking, you would know this already, thus the time and effort required to properly find tighter timings might have absolutely ZERO realworld benefit.