6700k high temps with H75

lessthanzach

Commendable
May 28, 2016
23
0
1,510
Hey guys. I've checked a few other threads and it seems like I am possibly alone in this and so I'm looking for advice. I upgraded my rig from an i5 3570k with an H75 (formerly was H50) that kept the temps below 80c even while stress testing for 30+ minutes. My rig now has an i7 6700k and the same H75 cooler. The cooler is maybe 2 years old at this point. I'm using HeGrease rated at 11.3mw conductivity and a very small amount.
qflGsy6.png
fMtk5JI.png

My temps seem to not be great at all. When first booted and let rest for a moment to idle, I can expect temps around 29-31c. If I run Aida64 for just about any amount of time, my temps jump into the high 80s. They even went as high as low 90s. Since this happened, I reduced the multiplier to 42 and reapplied the paste and still get temps in high 80s after a minute of running.
Am I missing something here? It seems like people are getting mid 70c temps using Evo 212 air coolers, I should be able to achieve similar temps with an H75 I imagine. Would the radiator need new liquid? Is this CPU just too much for this cooler? Also, I'm an overclock newb (last motherboard, just set xmp and bumped my frequency up to 4.2 and everything was sweet) and I'm concerned maybe the motherboard is giving too much vcore causing my crap temps - could that be something to look at? I'm really all ears, guys and gals. Wanting to get this hanging in the 70s under stress with maybe a slight overclock but even that isn't necessary if I'll need a better cooler.
Thanks in advance and let me know if I can provide more info to help.
 
Solution
Here's how mine is configured. It could certainly be tweaked a great deal more, I simply haven't had time, but it's rock solid for me as is and has been tested in Prime95 v26.6, Intel burn test, x264 stress test and Heavy load with no issues. Since you're only targeting 4.2Ghz, you can likely use an even lower core voltage than mine, which is currently at 1.325v.


Attention to the fact that Turbo is turned off, Intel speed step is enabled, the FCLK frequency is set to the same as whatever you set your CPU core frequency to, CPU VCCIO and system agent voltage are both set to 1.1v, flex ratio override is disabled, hyperthreading is enabled not auto and many settings are currently left at their auto settings, which for you, is probably a...
Something isn't working right. While that's not the greatest cooler ever, it should certainly be capable of doing better than the stock cooler, and even with the stock cooler you should have temps below 70°C if you're not overclocking. Are you overclocking with this chip?

I'd say either the mount is off, with too little pressure, the pump is failing or failed or that H75 is simply not configured correctly. Whatever the issue is, it IS an issue. You're not outside of spec in regard to temps, but you should be much lower than that I'd think at only 4.2Ghz.

Is the radiator configured as an intake or exhaust and where have you located it on the case?

Actually, on second thought, it MIGHT be about right if you're on a full time OC of 4.2Ghz with that 120mm single width cooler depending on your voltage settings.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


So, the mount is tightened down /tight/ this time I made sure. Used a screw driver to take it just past thumb tight. Nothing changed there from my 3570k. HW monitor and other software shows the pump RPM at the constant 1470 something RPM I have it set to (max all times), while the fan I have set at a fairly aggressive curve. I tried setting the radiator fan to 100% all time and temps still rise. The radiator setup is an exhaust but my PC doesn't really generate much heat as my GPU is a blower design and my PSU never even turns a fan on. So ambients inside the case stay quite low. Also, the radiator is located on the top of the case blowing out.
I was trying to overclock and had it @ 4.6 but lowered it down to 4.2 when I saw the temps. However, it didn't seem to make much difference. I'm going to (sigh) reset my bios settings to default, tick XMP and reboot and retest.
I really appreciate your help looking at this. Also, I'm pretty sure those Vcore numbers I posted are ridiculous for this frequency so I'm concerned my Mobo is full-stupid or I'm full-stupid.
 
Voltage is way too high. My 6700k runs at a max voltage of 1.35v, I never break 65°C with a 4.5Ghz OC using a single fan 140mm Noctua air cooler NH-U14S with a slim heatsink. I'd try dropping the voltage down to 1.35 and see how stability is there. There is much to account for in the bios settings when overclocking on Skylake. MUCH. Using automatic everything is a bad idea too. I also prefer to leave Intel speed step enabled and in my advance power profile in the control panel, drop the minimum processor power state to 5%, but I do that AFTER making sure everything is within tolerance and is stable with ISS disabled and both min and max processor power profiles set to 100%.

If you haven't already, I'd make sure you have the very latest BIOS installed as well. Also, be sure to manually install the chipset and other drivers from the motherboard product page. Do not rely on the native Windows drivers as they are not tailored specifically for the settings on your particular, or any, motherboard.

 

lessthanzach

Commendable
May 28, 2016
23
0
1,510

Nope. In fact, I just reset the bios to optimized defaults, ticked on XMP and that is my default Vcore. This is so frustrating, I guess I'll try and manually lower it.
 


Yeah, that's exactly what I was thinking after I took a second look at that. I failed to pay attention to the Vcore setting the first time I looked at it.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


Thanks again. To answer the prior thought of "if I'm constant overclock" - I'm not. I set min cpu state to 0% and it drops the temps almost immediately as soon as stress test is stopped. The temps are also erratic. They go from 67 to 89 to 72 to 85 to 65 again in a matter of 5 seconds. I'm on the latest F7 non-beta bios Gigabyte posted for my board (Gigabyte z170x Gaming 7). It seems like the bios is applying too much vcore. I'll try lowering it and see if it helps. At this rate, I'll have to turn off Intel Turboboost to not worry about frying my new gear.

Edit: After resetting my bios to "optimized defaults" and ticking the XMP to get my 3400 mhz ram to spec, I rebooted and ran a very quick test. Not looking good at all.
qF17Igf.png


Edit 2: Just manually set my Vcore to 1.35 and went into HWmonitor - It's still going over 1.4! I don't get why! This was way easier on 3570k, haha! At this point, I'll settle to get rid of all overclocks and just get my power and temp metrics somewhere I'm not stressed out about.
 
You don't ever want minimum CPU state below 5%.

I have the same CPU, and the Gigabyte Gaming 5 board which is identical to yours aside from a few SATA ports and some minor differences. I'll take some BIOS screenshots for you tomorrow so you can try out setting them to the same as mine, which were refined after a good deal of testing. Should get you close to where you need to be. Can't do it tonight as I'm beat and need to crash. In the meantime, I'd absolutely drop your CPU core voltage down to 1.35v and see how it does. I'm fairly certain that's where all your excessive heat is coming from.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


Dude, you're awesome. I feel bad bugging people about this as I usually just lurk and learn on my own. I appreciate your help greatly. I'll do my best to figure out why the Vcore isn't dropping even when I have it set in BIOS to 1.35 (still says over 1.4 in HW). Then I'll leave it be until tomorrow. I'm off work @ 6pm PST tomorrow so don't think I'm ignoring if I don't respond right away. Thanks again, what a cool community this is.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


Already did. There are 2 beta bioses that Gigabyte has up, but they only seem to "improve DRAM compatibility" and I felt that it was an unnecessary risk considering my RAM seems to be performing fine @ 3400 mhz.

That said, I turned Load Line Calibration from "Auto" to "Normal" (there was also high) and that seems to have dropped the temperatures quite a bit. I just ran 5 min of AIDA and only one core (#1) hit 80c, the rest stayed fairly consistent in the 70s.
UdRySvj.png
 
your cpu vcore is still outlandishly high.

i mean insanely high. I mean all your voltages are just wrong. I have a similarly clocked haswell (which uses more power then skylake) and my voltages are much much less across the board. something is seriously wrong. either your HWMonitor is reading the system wrong or your cpu is not configured right in the bios.

some of my max voltages with prime95

VID: 1.165V
VCore: 1.25V
Package: 74W
clocks: 4.2ghz
Temps: 72C max

now granted, you have an i7 where as i'm sporting an i5, but that still doesn't explain the vast voltage issue.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


I agree, fully. My Mobo has some feature in it set to Auto I bet that is jacking everything up. Best bet is to wait until tomorrow when I can get some picks, as Gigabyte's BIOS is quite different from the Asus I had with my 3570k and I'm actually starting to pass out as I type. Probably a recipe for melting my card since it's hard to concentrate. I'll definitely be on after 6 tomorrow to get this sorted.
With the lowered Vcore (though not much lower) it already helped a lot. So I'm guessing that getting it to where it actually needs to be will be what it takes to get me on the temperature gravy train. THEN I can start overclocking my CPU a little bit.
 
The "improvements" listed for any given bios release don't ever include ALL the changes to the code or microcode. Just because it says "improve XMP compatibility" for example, does not mean that is the only concern addressed in the release. I've never had issues on any modern system after installing a beta bios, and you can generally bet that once it's in beta release it's been pretty thoroughly tested already and in some cases they never even bother to come back and change it's standing to a final release. I see boards two, three, five years old that still show the latest bios release as beta even though it's been a proven release for years. I'd update.

BTW, I did mention that it would be a good idea to do so quite a few posts up as well. It MIGHT not make any difference in what's going on, but it might. Before the F5 bios release for the Gaming 5, I was having some issues with stability unrelated to XMP compatibility, and a few other glitches, but after updating, all those issues were gone. Just as an example.

Also, if a bios release makes changes to the memory settings, that can significantly affect CPU performance and thermals because the memory controller is on the CPU and will be affected by those changes. I'll get those screenshots for you and be back.

 
Here's how mine is configured. It could certainly be tweaked a great deal more, I simply haven't had time, but it's rock solid for me as is and has been tested in Prime95 v26.6, Intel burn test, x264 stress test and Heavy load with no issues. Since you're only targeting 4.2Ghz, you can likely use an even lower core voltage than mine, which is currently at 1.325v.


Attention to the fact that Turbo is turned off, Intel speed step is enabled, the FCLK frequency is set to the same as whatever you set your CPU core frequency to, CPU VCCIO and system agent voltage are both set to 1.1v, flex ratio override is disabled, hyperthreading is enabled not auto and many settings are currently left at their auto settings, which for you, is probably a good idea unless you specifically see it set to something else. Memory settings for you might be different, and if it works at automatic XMP profile settings, that's great, mine will not and I had to change the voltage values to not get reboot errors. I can't remember if this problem went away after the latest bios, so play around and see what you get when it comes to the memory. Anything that instantly doesn't allow it to POST has a high probability of being related to the memory settings.

148hcew.jpg



1snkv4.jpg



24q2xxe.jpg



5n6dm8.jpg



14y71wg.jpg



21jd1tk.jpg
 
Solution

lessthanzach

Commendable
May 28, 2016
23
0
1,510


Ok, cool. I meant no disrespect by voicing my concerns over the beta, just didn't realize it would contain other stuff. At work currently but I'll do that as soon as I get home. It seems even though I'm lowering VCore and switched LLC to standard something is still raising those values. Maybe the update will fix it. I'll report back.
 
I'd leave the LLC on automatic. Vdroop can be a worse or just as bad problem as thermals. LLC shouldn't be enough of a problem on the auto setting to cause issues if everything else is within tolerance.

You also have to make sure you're making the change to voltage core value in the CPU voltage settings, not on the main page. Be sure to hit enter after making a change as well.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


I am at my wit's end. I tried your numbers, no boot. Would get a blue screen with Hardware_exception_check. Messed with the numbers a bit by increasing the voltage to DRAM by a small amount seeing as I use 3400mhz ram. Still no boot. Then I'd bump the voltage a bit to CPU, would start to boot then give a bluescreen with a "clock watchdog timeout" error. Lowered the CPU multiplier to 44. Finally booted, waited for things to settle, then as soon as I opened a program, got the watchdog error again.
Tried disabling XMP with no luck. Tried turning DRAM voltage back to auto, no luck. Tried increasing Vcore up to 1.4, no luck. Only thing that worked was clearing the CMOS to go back to default settings. I am now booted and temps are staying at mid to high 70s while running AIDA at default settings (boosting to 4.2).
I also did upgrade to the latest beta BIOS per your instruction. I'm wondering honestly if I got a bad CPU - maybe one that wasn't lidded very well causing it to not cool efficiently or something.
I suppose now that I'm back to stock, I'll try bumping the frequencies a touch until it won't boot, then add Vcore until it boots and see if I can grab a 4.5ghz clock. If I can, then I'll enable XMP. My guess, as soon as I start touching frequencies, it's going to crap the bed again.
 
Try getting your CPU where you want it with the memory at 2133mhz and stock voltage which depending on your modules is probably somewhere between 1.2 and 1.275v. You can run CPU-Z and look at the SPD configurations on the SPD tab to verify. Just click on the dropbox for one of the slots that your memory is installed in and there should be about four SPD profiles listed. Voltage will be shown at the bottom of the SPD profile listing. One of the 1066mhz profiles will be 2133mhz, as it's double data rate memory so you multiply it times two. The last profile will probably show 1700mhz, and that is the XMP OC profile.

This way we can fairly well eliminate the RAM as being the source of the issue. Too hard to try and configure both an overclock on the CPU and memory at the same time. Get the CPU stable where it needs to be and then configure the memory.
 

lessthanzach

Commendable
May 28, 2016
23
0
1,510


I'll do that now. I currently have a 4.4 overclock with memory at lowest 2133 (stock?) value. Had to increase Vcore up to like 1.35 to boot at that though. Which seems odd. Starting to wonder if I have a really crap CPU that just isn't going to overclock. I bought this as a 5 or so year upgrade so I'm going to be slightly bummed if 4.4 @ 1.35v is going to be my max. That said, I did run AIDA for a bit and stayed under 80 just barely for about 5 minutes so it seems good in that regard.
1.2 is my stock RAM voltage and 1.35V is 3400 according to CPU-Z.
So, now that I'm at 4.4 (4.5 wouldn't boot - Hardware_Exception crashes) should I just XMP the ram and set the voltage to 1.35? Also, the other values you showed (VCCIO, etc) that were set to 1.1v, should those get bumped up at all if my Vcore is set higher than yours?
Finally, I'm pretty sure I hate AIO coolers now. Checked some comparisons and I am thinking I might have to get that Noctua. Not breaking 65c at 4.5 is nuts. My concern is I have a Deep Silence 1 case and I hate noise. These cases aren't exceptional at airflow due to their quiet design. Does that cooler require a really open airflow for the case?
 
If you're overclocking, quiet should be the least of your concerns. :)

Seriously though, lack of airflow through the case is probably a huge part of your problem. Silent type cases are meant to be paired with configurations that are intended to be overclocked to any significant degree. They are meant to be, well, quiet, which usually means dampened panels and minimal airflow through the case. You can't keep that radiator cooling well in an exhaust configuration without a high amount of airflow through the case. If you remove the hard drive cages you can mount the radiator up front, in an intake configuration, and it will likely run much cooler. I'd probably try that first.

Aside from that, my NH-U14S is nearly silent even after modifying the front case panel to open it up with a top to bottom grill, but as with any cooler, if you're overclocking, you need sufficient airflow for it to do it's job. That doesn't mean it has to be loud, there are plenty of very good high performance fans with low decibel levels. I'm not familiar with those Nanoxia fans so I can't comment on them, but Noctua, BeQuiet, Thermalright, Black noise and a few others make very quiet high performance fans that you won't hear except under rather high sustained loads.


This is my setup. I would think your case would work fine with this cooler as long as you have sufficient case fans installed. Seems like that case comes with two front intake fans and a rear exhaust. You'd probably want to add another top exhaust in addition to those three fans.

10egh35.jpg



hvxz74.jpg


 

lessthanzach

Commendable
May 28, 2016
23
0
1,510
That looks amazing. I've got my DS1 in white (hence my choice of Mobo) and I'm quite impressed with it actually. Everything I've done is to keep minimal temps inside the case while keeping noise down. The two fans up front can run at about 70% @ low noise and same for the 140mm in the back. My radiator is bolted to the top with a 120mm Noctua SP which is silent until CPU load kicks it up past 80%. My GPU is a reference EVGA 980ti so that it blows the air outside the case rather than vent inside. If I shoot the inside of my case with my thermal gun anywhere that isn't a motherboard I get 29c which tells me overall temps inside the case are good. That said, I increased my front fans to 90% and they're still barely audible so that should give me more positive pressure for when the radiator does need to come on.
I guess if I went with that cooler I'd point it at the rear 140 and my noise would probably be as quiet as the h75. Guru 3D pit it against an H100i and it smoked everything. Guru
Anyway, I've got her running good @ 1.325V @ 4.4ghz. I'll turn off Turboboost and tick that XMP profile and see what happens.
Thanks again for your help. Nothing like that sinking feeling when you realize maybe you should have stayed with what worked (3570k @ 4.2).
 
Oh, I can assure you that your Skylake system at 4.4Ghz fairly smokes that 3570k, even at 4.2Ghz, without having to even try much. That Ivy bridge chip was good for it's time, relatively speaking, but Skylake is a much better platform with far more potential.

Definitely turn off turbo boost. With a full time overclock, there is no need for it and it will absolutely steal control of your voltage and fluctuate it past where you want it to be, not to mention destabilize what might be an otherwise solid overclock by automagically configuring settings on the fly.