so after messing Im presently at the timings below , all other timings on auto and Tras and tRC were originally auto found , with this CH7 I am having random crashes at the minute if i push higher in game, with those i can use it fine but technically im still getting the odd shutdown ,only when i leave it idly crunching ,odd ,im reinstalling a fresh fresh wiondows , also, I have had most luck on auto settings but upping the current allowance to the memory and Soc and obviously running the memory at 1.35V , I find memory a laborious and individualistic thing to truly optimise these days(ie oc) since integrated memory controller heat is normally imho the weakest link of the effort and finding working settings can take time and a lot of testing but if i advance on these ill let you know..
the kit is corsair VengeanceRGB 16GB ddr4 CMR16GX4M2C3000C15
Controller temps can be a problem, but I would probably still play around with voltage. You still have some DRAM voltage to spare. It's not uncommon to hit a barrier where 1.35 isn't enough and anywhere from 1.355-1.45 becomes beneficial... ...usually at the point where timings are as tight as you can get them and speed is getting up there (seems speed increases nessesitate voltage increases more than timing decrease do, though.) Varies depending on type, but generally if you're trying to push the limits of your DDR4/mobo combination, there's a good chance you might have to pass 1.35. Really top-binned modules can do it, but you can't always expect it. You may just be hitting that point.
And again, a little goes such a long way. You're right, it's absurdly laborious... ...sometimes .005-.01v is the difference between stable and the occasional crash.
IME SOC voltage can be touchy, too. 1.1 isn't enough for my setup. 1.125 is perfect. 1.13+ and it goes to shit again. LLC can be just as important. I think of it this way... ...SOC needs a specific amount of power. So you can't run it loose and up the voltage so that it droops to the level needed, because then there will be times when it also goes just high enough to throw it off. It's not like a CPU where as long as the droop isn't too low, it'll generally work.
Reminds me, sometimes CPU voltage is a factor, too. In the past, upping it a tick has been the difference between working and not. Try it, if you can with the way you're running your CPU.
I dunno, could be a lot of things in your case, but personally if I have a config that boots and passes basic stress tests, but periodically crashes, then my next step is to see if I can't better balance the power to it. There's probably a specific scenario under which it crashes every time, while staying fine the rest of the time. This can either be a power need/delivery fluctuation or sometimes one particular timing causing it to crash during one specific operation that pops up when doing certain things. Somehow it's stumbling only under certain demands. I'd rather find out it's power first, before bumping timings tediously for nothing. Doesn't take nearly as long to rule-out.
Be interested to know if/how you get it figured out. These days I'm trying to learn as much as I can about RAM overclocking. There is much to learn from the suffering of others, so they say.
I've been playing around more with more modest overclocks on my 2600. Just wanna see how cool and quiet I can get things without taking a big performance hit. Today I managed
4ghz at...
1.13v. Now, I don't know if people are somehow over-volting like crazy with their overclocks or what, because I feel like every time I've read about someone trying to hit 4ghz with a 1600 or 2600, the voltage is anywhere from 1.3v to 1.4+v. 1.13v just seems insanely low. I found the instability line just below it. Low 1.12 is total lock-up on linpack... ...like before completing 1. High 1.12 is borderline... ...passes linpacks 9 times out of 10, but randomly shuts every few hours.
But it really has been completely stable at 1.13, and I don't expect it to fail, based on my experience with this particular chip. I did a couple hours of P95. 30 4gb linpacks. Another hour of RealBench. Obviously, I should do some longer tests - and I will put it through a lot of Realbench and maybe even a long stretch of small FFT's, but that it can pass thses suggests that it's likely going to continue working for me.
Maybe golden chip? I can tell you that going up every 50-100mhz past this point puts voltages closer to what you usually see, but it's still low. 1.23 for 4.1ghz. 1.29 for 4.2ghz. Those two I have already used and stressed enough to know they're completely stable. Only real explanation I see is that my sensors read low... ...but that would be scary low, wouldn't it? I'd hope not, but I can't rule it out unless I take my own readings. Imagine if it was actually .2v higher!
geez...
One thing is for sure... ...temperatures line up with the voltage. 64C for burner-type stress tests. 55C max for games, and a rare max at that. 50C is more typical. If I kept my fan curves up at where they were for 4.2ghz, I might see it go down another degree or two. Cooling setup in my specs. Standard air setup.
I might just keep it this way. Benchmark scores suffer marginally, but for my usage, I don't notice a difference at all. What I do notice is how I can't hear my fans at all. I'm not used to that. It's nice. Still performs a lot better than that stock boost, where sometimes it'll hit 3.9 under lesser loads for the faster cores and 3.8 across the board, but all cores average at 3.7 under real loads. Much bigger jump than 4ghz is to 4.2ghz, with the benefit of significantly lower power demands and great thermals.
If this pans out, maybe I'll try RAM overclocking again... ...see if running a little further from this chip's clock speed limit helps improve RAM OC stability a little bit. I have voltage headroom now to lock in stability with a slight overvolt. I can probably even lower CPU LLC and switching freqency... ...take a little noise out of the equation.
lol Maybe i'll raise vCore to 1.4v again, SoC - will leave it @ 0.987v for now. LLCs @ High right now. Will do it in 8 or 9 hours from now though, gotta sleep - i do NOT want to pass out from depriving myself from sleep in a middle of a test, now do i? 
jk Cheers.
