Thanks for your help Shilohh!
Well as soon as I touch the RAM settings the system won't post. I've tried leaving the CPU multiplier at auto and just getting the RAM to boot at 2133MHz giving more VCORE and VCCSA but no luck. When I just use the multiplier (44 or 45) and leave the RAM at 1333Mhz and the rest in auto it runs fine, but I'm not getting all the performance I could get out of the RAM.
DSDT or SSDT issue?
Listo
No. If you can't POST, it's a hardware problem. All CPU samples are different. If your 4930K can't run 2133 with given voltages but someone else's can, it's probably because your CPU has a weaker Integrated Memory Controller. It could also be a problem with the motherboard or ram or even just the BIOS settings. I found that I can't change "DIGI + Power Control > CPU Current Capability" to anything but "Auto" on my R4BE but I could on my R4E. If I select 120%, I lose stability and panic while OS X is loading even though my VCORE was set way under 120%. The point is that any 1 setting could throw things off so only make 1 change at a time before testing it.
Intel only warrants that the CPU will work at stock frequencies but they leave the K series unlocked so that you may see if you can get it to go farther. Ram manufacturers only warranty that the ram will run at XMP overclocked frequencies with the voltages and timings dictated by the XMP profile, but they usually specify a VCCSA much higher than intel or Asus recommend. They don't have to warranty the CPU life, just their ram. You can probably run VCCSA and VTT much lower if your CPU's IMC is strong. Higher voltage = higher heat, which = faster degradation, but more stability at higher frequencies.
Clone your boot drive before proceeding. An unstable OC can corrupt data (like your boot drive's file system). The machine can then act like it's unstable (like popping off while the OS is loading) when it actually could be stable. Fixing a corrupt file system can be tough if not impossible.
I recommend that you do a CMOS reset, "Load Optimized Defaults" and choose the basic BIOS settings outlined in the guide. Save this as a OC profile that you can load and revert to easily. I also recommend hooking up your Asus "OC Panel" or "OC Key", if you have it, to monitor voltages because HWMonitor currently shows 1/2 of actual vcore for "CPU Core" and "Memory Modules" is definitely not accurate. "CPU Core" temps seem to be accurate but but "Ambient" is actually the approximate CPU TCase or "CPU Temperature" in the BIOS's "Temperature Monitor". Start by setting only the XMP profile for your RAM in the "Ai Overclock Tuner" and stress test. For my stress test, I run prime95 torture test customized to use 59000MB of RAM (Because I have a 64GB kit. Leave about 5-6GB free for the OS). If you can't POST or run stress tests for at least 5 or 10 min without errors or crashing, try 1866 or 1600. Once you've settled on a frequency, record the voltages and timings specified by the XMP profile in SPD info on the Tool page. Switch from XMP to manual and set those voltages and timings manually. Stress again. If stable, drop VCCSA and VTT to 1.15 volts and stress again. If you crash or get errors, raise them by .010 volts. Repeat until you find the lowest stable voltage at that frequency. NOTE that Asus recommends that you stay below 1.200 volts for VCCSA to avoid damaging the integrated memory controller (IMC). I'm stable at 1.15 for for VCCA and VTT at 2133.
Once your RAM OC is set and stable, move on to CPU multiplier and VCORE. Start at a VCORE of 1.29 and see how hot your CPU cores get while running prime95 torture test for up to 20 min. If you get to 80c or above, I'd back off the vcore until you reach a max temp that you are comfortable with. Intel specifies a max Tcase of 66.8c. Tcase (Case Temperature) is the maximum temperature allowed at the processor Integrated Heat Spreader (IHS) and is usually 5-15c cooler than Core temps. I shoot for a average max temp of 75c (Core temps) with very short spikes up to 79c. Next, see how high you can raise the multiplier before you crash or get errors on the workers in prime95, then drop the multiplier by 1 increment. Once you've got that, you can now try to match your VCORE voltage and max core temps with offset instead of manual VCORE at full load (or just stick with manual if you prefer). Offset allows your vcore to throttle down with less load so your CPU can run cooler and use less power based on less load.
Last do a long stress test. For my long stress test, I run prime95 torture test customized to use all but 6000MB of your RAM and Unigen Heaven simultaneously for about 3 hours. Then I do a overnight run of memtest86 (NOT the older mac app but the new UEFI bootable USB stick) on multicore for 2 passes. Ignore a small amount of errors on the New Row Hammer Test as you probably won't find any high speed, high capacity, non-ECC ram kit that will pass this test without at least a couple errors.
