I only have a little less than two more hours left in which to help you today. Otherwise, see you tomorrow.

You are in good hands with HW 🙂

In the first place, using XMP can be problematic when you are OCing. When your system is 100% stable, it works fine. But when it is just a little unstable, it can not only complicate the diagnosis, but cause RAM failure as well. You see, the board does memory settings "training" on some successive reboots (you will see a few that take longer), and if you are OCing, and have an unstable setting when this happens, it can malfunction during this "training" cycle; either setting your memory incorrectly (the "auto" settings), or causing a failed POST attempt, or some other malfunction that I am not aware of.

You need to set your Ai overclock tuner to "manual" and manually set all those values for DRAM V, Memory speed, and all your primary latencies (be careful, memory speed will change when you change the "Ai overclock tuner" setting). This may or may not be part of your problem, but we need to eliminate all possible failure possibilities. But first, I see a discrepancy in your RAM and the settings the board picked: Your RAM specs are 9-10-11-27 (double-checked W/Corsiar site, assuming pt#CMT16GX3M4X2133C9) and according to the list above your XMP settings have chosen 9-11-10-27! Perhaps you inverted the RAS to CAS Delay (tRCD) and RAS Pre time (tRP)? You should double-check it before we continue. If your list is correct, then we should perhaps try 9-11-11-27 (it won't hurt anything, just improve stability). Personally, I like to use either AIDA64 or CPU Tweaker to read the XMP tables and set your secondary latencies too, but some ppl consider that just a little obsessive-compulsive so I'll leave that totally up to you. You could use the same numbers that you stated in your report above for your secondary timings, but I would trust what you read W/AIDA64 or CPU Tweaker more.

While we are discussing memory, I should also mention 2133 is a very fast speed. It's possible that your CPU IMC will not do that speed stable W/your CPU OC. You may have to kick it down a notch in speed (or reduce your OC), but let's try 2133, fixing your other issues first.


The next thing I need to discuss with you is your static CPU Vcore OC strategy. There's nothing wrong with your static Vcore settings, and I see you have set your VCCSA to an expected number for your OC level too; but do you know it will take much more amperage, and thus $$ when idle like that? And your idle temps will not be low neither. I have been to 4.7 on variable, but Vreg throttling occurred during LinX; and at 4.6 stable (for a few months) without Vreg throttling - on variable Vcore - without sleep issues. I have C1E - C6 enabled, and I get great idle economy. That said, it is a personal preference and you can do what you want; I will help you with it either way. If you do not want to try variable, I will not push it.

On static Vcore, when you set 1.4Vcore you theoretically get 1.4Vcore (perhaps modified a bit by LLC, Anti surge support, and CPU current capability) no matter what your CPU is doing, and no matter what your current operating Turbo (or not) multipliers are. When you use variable, the Vcore decreases when the CPU is idle (at lower - 12 - multiplier). However, set wrong this can cause a sleep/idle issue. If you are going for max. benchmarking - competition level - OC, for bragging rights or whatever, then static is the way to go; but then you should also have extreme cooling W/a Vreg cooler too.


The first real non-memory CPU OC issue I see W/your settings is your current capabilities. U-haul truck engines used to have "governors" on their engines to prevent them from going over 45-50MPH. Leaving your current capabilities (and Anti surge support) at "auto" or 100% is the Asus equivalent to trying to run a race without disabling the governors. You can run the race, but you will never get over a certain speed because the governor is starving your engine.

Standard 100% is for stock voltages and speeds. And furthermore your RAM is not at what Intel calls stock speed, it is likely to need more Amperage ("A"); and your CPU and VCCSA DEFINITELY need more A. The "capability" settings are to prevent over-current damage in the event of a failure ("short-circuit") in the chips or boards. But that event it is catastrophic anyway, doesn't really help much. You need to set:
CPU current capability: 130%
VCCSA current capability: 130% minimum
DRAM-AB current capability: 130% minimum
DRAM-CD current capability: 130% minimum
and I recommend:
DRAM-AB power phase control: Optimized
DRAM-CD power phase control: Optimized
He talks about this in the (obsolete) Asus ROG YouTube video OCing the RIVE 1/3 (they never made 2 & 3 as far as I can tell). Don't rely on that vid too much though, it was made when the RIVE was first released, with the original release BIOS and memory designed for the X58 chipset (the older version XMP tables).


The next issue is "Anti surge support." This means when your CPU A goes over a regular value (for stock speed/Vcore), your Vcore is throttled to reduce the A. Disable it. You will then find you can reach higher OC easier. This also could be an issue with coming out of sleep, when your CPU starts with static Vcore the A could spike and Anti surge support could be throttling the Vcore at that moment.


Minor issues: "Rampage tweak" mode 1 is for compatibility. Mode 2 is for "performance." Auto may choose mode 1, IDK, but I would suggest you set it mode 1 to be sure. This is just another variable to eliminate from your issues. AFAIK "performance" is really the mode for those who do extreme memory tweaking (using an oscilloscope, etc.), not us regular ppl; or maybe once you get your max. OC stable you could try it (I don't recommend it). Perhaps it has something to do with how agressive the RIVE memory "training" cycles are. Using the compatibility setting (mode 1) may reduce BSODs for you.

VTT. You have VTT on manual? That's OK it just looks a little low. Just keep in mind raising it may help your OC, only testing will tell. If it works that low, fantastic.

CPU load-line calibration ("LLC"): I'm really not able to help much with this setting in static Vcore mode, but in variable Vcore you would want it set to "medium." This would prevent your Vcore from dropping during peak loads, and allow it to reduce to a nice low value at idle - when using variable Vcore. Perhaps someone else can chime in about it's proper setting when using static Vcore.

HiVizMan just PMd me (answering PM):
He said go with medium or high LLC. I suggest medium for starters.