Corsair H115i high temps.

[EpIckFa1LJoN]

So I started overclocking my CPU yesterday and I feel as though the thermal wall prevented me from going any higher. As of right now gaming loads are pretty good. I have a 4.6GHz overclock on my 6700k, and the voltage is pretty average I believe. It is set to 1.305V in the BIOS but it will vary, highest I saw it go when tuning yesterday was 1.338V (at 4.7GHz I think) so I think I have a little room in that area. But when I set it to 4.7GHz my temps hit 87C and the same core failed Prime 95 short FFT test twice in a row (even though it passed LinX twice) before I set it back to 4.6GHz. So now my temps will max out at about 83C (even though gaming loads never come anywhere near that) which is about the thermal limit most people give for the 6700k (or at least the safe limit).


So now the cooler. I feel like those temps are really high. I might be crazy but it seems like a cooler that cools as low as 23C on idle should be a lot better. It cools pretty fast as well as soon as the tests are over it goes down to 50C in about a second and goes down to 27C after 4-5 seconds of being on idle. Gaming load temps are in the high 40s low 50s.

So with anyone else's experience are these normal temps? or should I try reapplying the thermal paste or changing the fan orientation, kicking on the fans/pump more?

Fan details:
All case fans: Corsair ML140 (set to 1100RPM for near silent operation)
Cooler Fans set to 68% in Corsair Link (hovers around 1250RPM also for near silent operation)
Fan pump set to *performance* mode in Corsair Link
Fans set to exhaust (as opposed to the recommended intake)
Thermal Paste: Arctic MX-4 (may need to be reapplied, I took the water block off to move the motherboard a few months ago and I never reapplied the paste, I didn't know if I was supposed to or not and didn't want to deal with the hassle)


Any help would be greatly appreciated.

 

[hwlr77]

if you took your heatsink of the cpu and didn't clean off the old thermal paste, and reapply new stuff then, that should be the first thing you do.

 

[EpIckFa1LJoN]

if you took your heatsink of the cpu and didn't clean off the old thermal paste, and reapply new stuff then, that should be the first thing you do.

Well that's not what I wanted to hear but okay xD.
It's only a 30 minute or so job but i'm pretty lazy, lol.

From researching it a bit it looks like it's about time anyways, so I'll do it this afternoon and let you know how it goes lol.

 

[Maebius]

Hello!
After you're done reapplying your paste, note a few things!
If you're using a rather new prime95 version, don't....use IntelBurnTest instead as prime versions over 26.6(?) produce completely unrealistic heat due to the AVX instructions, unless you disable them.
I have a corsair h105 which is a bit worse than yours and with my 7700k, 5GHz and 1.3 vcore I get max 75-77 degrees on IBT, 50-65ish on games that are cpu intensive (like ME:Andromeda, that thing is terrible!), in ambient temps of 25-30 degrees...(summer is upon me!)
When you say that your H115i is exhaust, does it pull air from the radiator or does it push air through the radiator and out of the case?
Depending on your case's airflow, the intake solution is always the best for the CPU (but worse for your GPU and the rest of your components).
As I tested mine, blowing air through the radiator and out (up) of my case proved the best all-around option as I had a big enough case where the GPU's hot air didn't change the internal temperature drasticaly
If you're not too bored, you can try it out.... as a reference, I had 5-7 degrees higher max temperature when it my radiator fans where set as pull-exhaust (up).
Also, I believe even at 87degrees, it shouldn't fail so it's a voltage (due to vdroop or not) , memory issue, or both.

 

[EpIckFa1LJoN]

Hello!
After you're done reapplying your paste, note a few things!
If you're using a rather new prime95 version, don't....use IntelBurnTest instead as prime versions over 26.6(?) produce completely unrealistic heat due to the AVX instructions, unless you disable them.
I have a corsair h105 which is a bit worse than yours and with my 7700k, 5GHz and 1.3 vcore I get max 75-77 degrees on IBT, 50-65ish on games that are cpu intensive (like ME:Andromeda, that thing is terrible!), in ambient temps of 25-30 degrees...(summer is upon me!)
When you say that your H115i is exhaust, does it pull air from the radiator or does it push air through the radiator and out of the case?
Depending on your case's airflow, the intake solution is always the best for the CPU (but worse for your GPU and the rest of your components).
As I tested mine, blowing air through the radiator and out (up) of my case proved the best all-around option as I had a big enough case where the GPU's hot air didn't change the internal temperature drasticaly
If you're not too bored, you can try it out.... as a reference, I had 5-7 degrees higher max temperature when it my radiator fans where set as pull-exhaust (up).
Also, I believe even at 87degrees, it shouldn't fail so it's a voltage (due to vdroop or not) , memory issue, or both.

Thanks for the reply, that's some great insight. I was already concerned the voltage may be too low, since it was set to 1.305 and I see others with much higher values for similar overclocks (and is set to adaptive mode. I am using Prime 95 version 28.0, in addition to LinX. I do notice when I run a specific Prime 95 Test (1344 Max) It gives me much lower temps, and in gaming I never even come anywhere close to 80, but I failed Prime 95 so I lowered the multiplier by 1 and stopped pushing (even though I feel I can go much higher, 1.338V max at 4.6 and 100% stable seems pretty dang good from what I've seen). The fans are set to exhaust case air out the top (recommended by Corsair is to pull cold air from the top and into the case so I will probably try that). I have all SSD drives so there isn't much heat in the case from that, but I do have a Zotac 1080 Ti which can get pretty hot, but luckily it is set to silent mode so I can can always up the fan curve a bit to compensate.

I really want to push the chip to a still safe thermal and voltage limit and get the most out of it.

So what stress tests do you suggest I use? (doing 30 min of 1344 Min/Max FFT, Run in place, 15min each, and roughly 10min Small FFTs, and two LinX tests each time I tweak.)

 

[Maebius]

I use IntelBurnTest , AsusRealbench for initial stability testing.
If it is stable for an hour, it's usually ok.

After that I always do some real 4+ hour encoding or file conversion (like WinAvi to convert an mkv to avi).

Then I do a couple of 3d Mark Firestrike runs but I've actually never ran into problems there, if everything else has passed...it's mostly to see the progress while bumping up the frequency.

Obviously there's nothing like the real thing so playing proper, cpu demanding games like BF1 or ME:Andromeda will sooner or later crash/hang if the system isn't stable.

What LLC setting do you use while overclocking? If you let it on auto it's usually not aggressive enough and the voltage drops quite a bit under load, making it unstable... (in Asus motherboards 1 is the lowest and 7 the highest).
For instance on auto I had set vcore ov 1.31 in bios and under stressload it actualy went down to about 1.27.
With LLC 5 and vcore set at bios to 1.305 it drops to just 1.296, making it perfectly stable.... of course not every chip is the same but you get the idea.

If you flip your radiator/fans over to work as an intake (didn't see your pc's picture as I was at work), your CPU will most likely run a bit cooler.
Overall, your inner stuff will probably be a bit hotter as you'll have 2+2 intake fans and just one exhaust...with the exhaust being right next to an intake... dunno airflow-wise how that will work, but try it out...like a formula (motherboard name pun intended hehe) windtunnel test.

Cheers o/

 

[EpIckFa1LJoN]

Haven't changed too many of the settings it's pretty straightforward so far. Using the guide for the same UEFI BIOS on Asus' site. LLC settings isn't on there to my knowledge, and in fact every guide I see I have to look because apparently Asus has a different term for everything. VCORE or multiplier is nowhere to be found in the extreme tweaker page. I haven't noticed the Voltage dropping under my limit though. I believe that is how it is working it will go over but never under my set voltage. Further testing is needed to make sure.

I was going to use AsusRealBench today and see as that is the one suggested by the Asus guide. As for gaming testing even on battlefield 1, the CPU usage doesn't get that high. It actually gets much higher on FO4 and on a single core in WoW (usage is at like 90%). I don't really have the patience or really the time to test everything that long so I rely on about 1 hour of the tests I listed with a little bit of gaming to be sure just to check it. Before it was stuttering pretty bad and it no longer does, and these are in gaming so I really assume the OC is stable, but I am getting hints as to the necessity to increase the voltage or decrease the temp as the 6700K does throttle at 80C and I noticed that when I hit 4.5 GHz the LinX test slowed down my entire system to a crawl and it would freeze a lot. So I have plenty more work to do but I can't really do 4+ hour tests or things like that. No time or patience lol.

 

[Maebius]

You should have LLC...it's CPU Load Line Calibration
It's either in external digi power ctrl or internal cpu power management subsection of the extreme tweaker menu... even on my old 2500k motherboard had LLC settings to play with (it had auto, low,med, high or something like that.... your board should have 1-7 or 1-9 numbers other than auto)

What are you using to watch your real vcore? Asus Rog's CPUz shows ViD (requested voltage) and not actual.
Use HWMonitor and look at VCore voltage number (not ViD) when stresstesting... if you don't get voltage drops then cool

The 6700k's supposedly don't throttle at 80c so dunno with that... it might be a memory issue even... if you're not, you should check your overclocking potential first with stock (2133GHz) memory frequency and not XMP.

I don't play FO4 anymore but in really big multiplayer maps BF1 was quite taxing (also:depends on the refresh rate you're playing at... if it's just 60 it's not too demanding). ME:Andromeda smells of terrible optimization...running it is almost like stresstesting the system hehe
WoW is peanuts, it only uses 1 core and that core spikes only on the start of a fight with bloodlust on and recount showing

 

[EpIckFa1LJoN]

It does have SVID, which is the external voltage controller. Something like that. Didn't see any other options. It was enabled, reccomended is disabled so I disabled it. Set voltage to 1.34v and cranked it back up to 47 on the multiplier and it passed Small FFT on Prime, So I guess it was the voltage.

I'm about to download Asus and see what happens. Reapplying the paste did almost nothing, so it looks like it wasn't that. I was getting about the same readings before and after. (maybe im not using enough??? I put a little smaller than a pea sized dot in the middle and then put on the block?? followed every direction i've seen to do that, idk)
Anyways with the voltage at 1.34 the spike temps are way higher. I got 96C on one core spike.

HOWEVER it does seem my (equilized) normal and idle temps are lower. idle im hitting like 23C at the same settings I was hitting almost 30C on idle with. and during the stress testing it lowered about the same interval [4-7C] (maybe the paste helped after all??)
[Thinking out loud here]

So anyways it seems stable so far at 4.7GHz gonna keep pushing then lower the voltage little by little and see what I can come up with.

UPDATE: Upped the multiplier to 48, LinX failed immediately so I went back to 47, started lowering the voltage. So far so good at 4.7GHz. RealBench max temp is 79C so far. Going to keep lowering after I get to the gym but it seems the thermal paste has helped as well as getting the voltage better. So far going in the right direction!