Originally Posted by Belial
you realize a PEA
is WAYYYY too much thermal paste, no wonder your temps are awful. You should be using only enough paste that it ends up spreading out to cover HALF of the base. This is because the die is where the heat comes from, not the IHS, and due to convexing.
This is what WAY too much paste looks like, like 2x too much:
This is a lot better:
This is about as perfect as you can get with application:
You need to use a small ricegrain amount, bordering on less rather than more. Seriously, take a look at a ricegrain, like --- much. Also, delidding, and, be aware that the hyper 212+ uses jet engine fans while the h110 uses quiet fans (still should blow h 212 out of the way but it should close the gap by a lot).
Max temp is 90-100C on Haswell (most go by 90-95C depending on how conservative you are), you have way omre headroom. Try to see if you can do email@example.com, and raise the multi up/down and the voltage up/down based on how much thermal headroom you got and how stable you are there. Delidding will allow you to get an extra 100-200mhz overclock too, at least.
Also, VRIN really affects overclocks, it basically lets you use like .1-.3 less vcore if you optimize it. I've personally found it best at ~.55v above vcore, on less vcore you could use less so assuming you stayed at firstname.lastname@example.org, you could probably lower vrin to like 1.5vrin which will really drop your temps.
After you dial your CPU, mess with uncore. It'll give a performance boost, not huge but not small either. Try to set it about 100-500mhz below your CPU clock, and use VRING to stabilize it, and dont go above 1.3vring (it doesnt add much heat but i hear 1.3vring is the max, im sure you could do more but i couldnt say that for certain).
Once you do that, go for RAM. I'td help if you told us what exact ram you have, dont just say 'gskill ripjaws 8gb' because there are like 50 different ripjaws, i mean the model number, the exact speed/timing ram you bought, etc. We could tell you how far it could likely go and what RAM ICs are in it (basically, there are only ~4-5 different RAM and a few sub-types, everything is rebranded spektek or qimonda (really crap ram), or hynix or samsung (MFR, BFR, CFR hynix and HYKO or HCH9 sammies). Or micron d9 (generally crap but crucial low profile ballistix are amazing micron's, it looks like yours is the regular crucials which are crap though).
Basically set your RAM to 1.8v, set all timings to auto, see the max frequency you can do, and then you just dial it back a step or two from your max frequency and tighten it up (you can get a lot more involved than this but this is the quick way to do it. ie set 1.8v, try 2000mhz, then 2133, 2200mhz, 2400mhz, oops, cant do that, so go with 2133 and tighten CL as far as it'll go, then go X / X+2 / X+1 / X+Y+Z and use Command Rate 1.
No one has ever killed their chip at 1.5v, I'm pretty sure you need 1.6-1.7v+ just like Ivy to kill your chip. That guy, was just an idiot, he also killed a 2nd chip on 1.2vcore. It's because he was using, god only knows why, 2.2VRIN, which we all know is danger levels. Why you would use 2.2VRIN on 1.2vcore when everything and everyone says use only +.4-.5VRIN above vcore, I don't know. Even on 1.5vcore that means you shouldn't ever go above 2VRIN, even 2.1VRIN.
On top of that, that guy also had a faulty motherboard with bent pins, it's very likely VRIn wasn't what killed his board but that he was simply shorting out his chips. Read his thread people, don't just read the title...
What fans are you using? What thermal paste? There is no way you have only 70-80C on a stock h110 on a non-delidded chip on 1.5v! My H110 with 4x 2200RPM Yate Loon highs, in an open air set-up, as intakes, delidded, hit ~79C max in non-avx stress tests. And that's email@example.com/2.03VRIN. I'd love to know how you got such great temps on such high voltage!
Unless you are using cinebench for stability and heat testing... cinebench is awful for stability testing, it's nowhere close to stable just if it can pass cinebench, wont pass 5 minutes of p95 and would be way over 100C, ha, but some non-AVX stress test like aida64 or h264 benchmark should push you way over 90C, even 100C, on a stock h110 without delid..
Anyone tried Intel Xtreme Tuning utility stress test? You would think that, given all the hubbub and issues with AVX and such with haswell, that maybe the official intel stress test would work really well at testing haswell? Im going to try it out, I didn't have any particularly special results with it on z77.