Core 1 Running 10+ °C hotter?

[revilo951]

Hi everyone, first post here. Lots of useful info found in the past.

I have a brand new system with an i7 6700k, an Asus Z170 Deluxe, an Asus GTX 970 Turbo, and a H110i.

Whether under load or not, core 1 seems to be running much hotter than the other cores. See screenshots.

If I get this properly sorted, I have something to give to to the most helpful person, for all the helpful info I've gotten from here in the past.

TIA,
Revilo951

 

[Darkbreeze]

Core 1 on the 6700k has a higher stock turbo clock configured than the other cores, if you check the bios individual turbo core settings. It's also more likely to be running any single core tasks. Mine is the same, with a slightly higher temperature than other cores. Even at "idle", there are often processes running that may not constitute significant loads.

That being said, it's also possible that your paste job has not been uniformly spread on the CPU lid and there is less thermal transfer occuring near the location of the #1 core. It might be worth investigating as I'm only seeing about a 5°C margin between core #1 and other cores, and I'm not seeing it all the time. 10°C could indicate an issue, but may not as I've indicated.

If you download and install HWinfo, boot into windows, run HWinfo, allow all the startup processes to finalize so that the core usage settles at pretty much zero for all cores and stays that way for a while, and you're still seeing a 10°C margin, I'd repaste and use about half of a #2 pencil eraser sized amount in the dead center of the CPU lid after cleaning both the CPU lid and heatsink using isopropyl alcohol.

 

[Golcsae13]

This is fairly normal depending on how well the thermal TIM was placed between the CPU die and IHS. You can try reapplying thermal paste and re-seating your cooler, but if the results remain the same then there isn't much you can do. As long as the cores remain under the thermal limit, you shouldn't be worried. You can also delid your CPU and apply the thermal paste yourself, but this is very risky.

 

[Darkbreeze]

You sure as heck don't want to try delidding an already weak Skylake CPU. Chances for failure are higher than they already were with previous models due to the thinner substrate. Bad idea, worse even than it used to be. I'm fairly sure it's normal core variance given that core 1 and it's hyperthread seem to most often be assigned the load on any single core processes, at least on my system. Or, as I said and you backed up, it might be an uneven paste job.

 

[Darkbreeze]

Here I have about a 14°C variance between core1 and other cores, with very little load. It's just that what load there is, is mostly running on that core and it's hyperthread. I know my paste job is even, as I did two previous samplings to test the spread pattern.

 

[revilo951]

at the moment, it has the pre-applied thermal paste from the H110i. I did have to remove and reapply the block twice, so that might be it. I'll try new paste as soon as I get a chance (Aussie here).

BTW, what is a decent recommended thermal paste?

 

[babernet_1]

Yes, you either have a bad heat spreader inside the CPU package that really you cannot do anything about because delidding is so dangerous. Or you could have a bad connection between the CPU and the heat sink. First check that the heat-sink is properly installed, all mounting screws in place and snug. If so then remove the heat-sink, clean it and the top of the CPU with alcohol, re-apply new past, and try again. Watch some youtube videos on this to see it first.

 

[Darkbreeze]

If you removed and reinstalled the water block, and did not clean and apply new paste, then it could well be the problem. Might not be. Might not even "be" a problem, so long as all temps on each core remain below threshold specs, but it certainly could be. The stock thermal TIM sucks anyhow.

I'd probably recommend using any of the premium TIM. I still like Arctic Silver 5, but mostly I use the Noctua NT-H1 or Cryorig Cryopaste. I don't like the liquid TIMs because they're messy and have results from very good, to extremely bad, depending on application. Any decent premium paste is fine so long as it's applied correctly.

And by correctly, I don't mean what ten idiots on Youtube tell you to do, or any of the extravagant methods listed by the various paste and cooler manufacturers. Unless you're using a specialized paste, like liquid metal or ceramic types, just put about half of a #2 pencil eraser or snow pea sized amount, or large uncooked rice grain (Think wild rice, not white rice), dead center on the CPU lid after cleaning, and install your block. As mentioned, make sure the waterblock is fastened evenly on all sides as well. Using incremental steps to tighten fasteners, moving from fastener to fastener in an X pattern, is recommended.

Methods shouldn't vary much if at all from what's indicated here, regardless of cooler heatsink or waterblock:

http://www.tomshardware.com/faq/id-2520482/solving-temperature-issues-hyper-212-evo.html

 

[revilo951]

I have done some more temperature testing, with and without overclock.

See screenshots here:

http://1drv.ms/1Q3UjML

 

[Darkbreeze]

What were you using to apply the load in that shot where the temp goes to 92 degrees? What is your fan profile set to? No way at only 4.5Ghz with an H110i you should have temps that high. I'd repaste to begin with and go from there using the methods outlined above.

My 6700k temps don't go over 63°C on any core using air, so there's definitely either something wrong with your cooling configuration, paste job or you're using an inappropriate test utility. Are you using Prime95 version 26.6 or something else?

Where is your radiator located and is it in an intake or exhaust configuration?

 

[revilo951]

Using prime95 v27 and v28 (not together 😉 )

I have a Silverstone KL05-W case, with 2 fans in the front (intake), one in the back (exhaust), and the radiator in the top (exhaust), with the fans in pull configuration. The rad is inside the case, and the fans are outside with the sheet metal of the case between them, like this:

I haven't got any isopropyl alcohol at the moment, so it will have to wait a little.

 

[Darkbreeze]

Using heated air, in an exhaust configuration, isn't very efficient, but considering it doesn't look like you can front mount that radiator I guess you don't have much choice in the matter. It however usually recommended to have the fans in a push configuration, which is why every manufacturer designs their heatsinks and radiators to be used that way. Plus, with the radiator inside, any radiant heat dissipation from the radiator stands a chance of remaining trapped inside the case, while if the radiator was where the fans are and the fans were where the radiator is, some additional exchange of heat would occur without it being trapped in the case.

Using cooler outside air in an intake configuration with radiators can make about a ten degree difference, so if that's an option it would probably help. If you run it with the side panel off it wouldn't matter, in which case, it doesn't. You might try that, at least for testing.

Still, that doesn't address the core variance we see. First thing I'd do though, is quit using those versions of Prime as they use AVX instructions which is not recommended for thermal testing. AVX instructions are not steady state, and steady state is the key to accurate thermal testing. AVX also creates unrealistic temperatures. Use version 26.6, here, and see what kind of results you get then. It may be there is no problem at all.


http://windows-downloads-center.blogspot.com/2011/04/prime95-266.html


Liquid cooling isn't one of my best areas, which is to say I have experience, but not extensive specialized experience like some members, in that area, so take it with a grain of salt, but those would be my initial recommendations at this point.

 

[revilo951]

OK, I have replaced the TIM with IC Diamond (man, that stuff is stiff!).

I also reversed the rad fan direction.

See screenshots here:

http://1drv.ms/1JNl1YL

Thoughts?

 

[babernet_1]

Face it, you have a bad IC. There is a bad thermal junction between the die and the top of the chip. Either live with it, delid, or get another one.

 

[revilo951]

Dang. Can I RMA that thing?

Or, could I under-clock that core a little? Is that crazy?

 

[babernet_1]

Dang. Can I RMA that thing?

Or, could I under-clock that core a little? Is that crazy?

No you can't RMA. All they guarantee is that it will run at 4G/4.2G boost and CAN be overclocked. You can't overclock a single core less than the others.

Really, only an issue if you really want to overclock a lot and run the chip hot. You have lost about 10/15 degrees of heat headroom. Not the worst thing that has ever happened to someone.

 

[Darkbreeze]

I don't see any of those screenshots showing 100% usage under a Prime load using v26.6, but if that is full load then I'd say you're still fine where you are. Real world use, you're probably, almost certainly, never going to see 100% usage on all cores for any sustained length of time anyhow, so 72 degrees under full load would be entirely acceptable, if a bit higher than what you really want. Considering the OC, it's not bad at all.

Now if that's only 24% usage, then it's a problem and I'd drop the clock a bit.

 

[revilo951]

That's interesting. Task Manager shows 100% usage, but realtemp only shows 24%.

What's going on, and who to believe??

 

[Darkbreeze]

HWinfo. Run Prime95 version 26.6 and HWinfo. Those are all you need to verify thermal compliance on any past or current chipset/cpu combination. Not HWmonitor, not Open hardware monitor, not CoreTemp (CoreTemp is good for quick reference though) and for certainly not task manager. Task manager is wrong more often than it's right, about a good many things.

 

[revilo951]

Check the hwinfo comparison screenshot in the folder.

http://1drv.ms/1JNl1YL

Also interesting to note the temperature differences (or lack thereof) with the rad fans in push vs pull.

 

[Darkbreeze]

Temps are hard to figure sometimes, as related to fan orientation. Did you swap the fan and radiator locations, or just undermount the fans? Is your side panel on or off?

As far as your thermals, how long had you been running Prime when you took that last screenshot that has the HWinfo sensor readings?

 

[revilo951]

I just turned the fans over, so it is blowing through the radiator into the case. Probably not the best thing for overall case temps but there is fresh cool air to cool the cpu.

It had been running for about 5 minutes. the temp wasn't increasing at all. it was jumping around by 2 or 3 degrees.
please excuse any mistakes - mobile keyboard.

 

[Darkbreeze]

No problem. For thermal testing, we like to see Prime run for 15 minutes. While temps may not seem to be moving at all at less than that, they may again begin to climb during specific tests. Each test is not the same and some are more demanding than others. Thermal complaince can pretty well be assured after 15 minutes, stability between 15 and 24hrs with 24hrs being much preferred as we've seen plenty of systems that we still error free at 15hrs but had worker stoppages somewhere between then and 24hrs.

I'd run it again, take another HWinfo screenshot at 15 minutes into the test, and post just the HWinfo sensors window with the temps. For what you've shown so far, while there IS a fairly large variance between the cores, the overall temps are still within spec. Let's see if that's still the case at 15 minutes.

IC diamond is particularly finicky about application. If you used less than about 3/4 of the size of a #2 pencil eraser you probably do not have full coverage. If you used more than the visible section of a new #2 pencil eraser, you almost certainly used too much. IC diamond recommends erring to the side of a little too much when applying compound, versus too little, but I'm not sure I agree. You might wish to pull your waterblock again and, while trying to not move it around which would skew the results, see what kind of coverage pattern you got.

I'd also double check that your mounting gear is all evenly attached to the motherboard before installing the waterblock. One fastener not tightened the same as the others could potentially create an uneven tension between the surfaces.

 

[revilo951]

Thanks for your help everyone, especially darkbreeze. I haven't got a heap of time on my hands at the moment, so I am happy with what I have at the moment, until I can spend some more time on it. Sorry I have been so long replying.

darkbreeze, I will PM you about gift mentioned in post 1.

 

[Darkbreeze]

Hey, no worries. If you have further questions or wish to discuss the issue further later, you know where to find us. Good luck.