I feel it's the timing and voltage of the ram which I've never been able to get quite right since there are so many variables.
I've had 4 sticks of A-Data 1600 dual channel timed at 9-9-9-24. 1.74 volts For years even though the stick says 1.65v it was never stable.
If changing these settings from the defaults resulted in performance that was "never was quite right" and "never stable", then why stick with them.
That should have been a sign that something was wrong, and you needed to back off. This clearly indicates to me you need to reset your BIOS and undo all
your clocking changes and run at the "expected" default settings.
Then see what happens.
Why did you use that RAM? When buying RAM (or a CPU) you should ALWAYS consult the QVL for your motherboard. You MUST buy a listed CPU and you MUST buy RAM with the same specs as listed RAM to ensure compatibility and support. Your Memory QVL is found here
. Note there is no DDR2 listed. And the specs
say DDR3 1.5V RAM is supported. DDR2 RAM will not even properly install on that board without using excessive force to jam the sticks in the slot - so I will assume you are using DDR3. But, I fear not totally compatible DDR3.
It sure looks to me like you selected the wrong RAM to start, and you are pushing it to unstable limits. Not a hardware problem, but, and no disrespect meant, but a user problem. :(
Finally, note this is a "hybrid" motherboard. It supports both dual-channel memory with 2 sticks and triple-channel memory with 3 sticks. If me, I would toss your current RAM into your spare parts bin and buy a triple-channel kit of 3x2Gb for 6Gb total (or 3x4Gb for 12Gb if budget allows) of RAM with same specs as the QVL listed RAM - since it appears
you are running 64-bit Windows (assuming the 4x2Gb spec is right).
FTR, this illustrates why I really don't like overclocking, except maybe
when using the motherboard maker's own dedicated OC tool (if supported by that board) that manages all the timings for us users. As you correctly noted, there are just too many variables that are difficult and take time to understand, even for experts, in order to do it right, and safely
- all for performance changes that typically are only truly noticeable on paper and benchmarking programs.