Warning: Some i7-6820HKs and i7-6700HQ have Uneven Core Temps due to Uneven Heatsink

It's fun, you know there are solutions, but which one is it?

After successfully fixing heat issues so many times with re-pasting and de-lidding, it sure was a surprise when those go to solutions wouldn't solve my 2 personal LGA CPU's that had internal failures.

What is surprising is that these "hot core" CPU's are found across a range of laptop models, makes, using different cooling systems with different methods of applying pressure evenly across the heat plate.

So far we haven't seen anyone come in saying they had good core differential temperatures *before* re-pasting, ending up with Core 0/2 differentials after re-pasting.

I would expect failures after re-pasting if they didn't get the pressure applied evenly, as manufacturered.

I would think that result from re-pasting would be very common if "hot core's" is heatplate pressure related.

It will be very helpful to get a "how to" apply pressure evenly to cure the "hot core" problem if it is indeed a mechanical problem with the heat plate pressure.

Very exciting

 

Great thread! I'm was deciding between which 6820HK laptop to buy but this is greatly discouraging... Will be following iUnlock's findings..

 

I agree and have a feeling that we'll get down to the middle of this here very soon. I'm would take my 17R4 apart, but there exist no issues there so it wouldn't be of much help. @Mobius 1 will have some post repaste data here very soon as will I so we'll keep everyone posted.

 

I've had this unevenly HS seating suspicion from the beginning.Am i the only one who see this larger gap on the right side?I screwed them in as much as i could...
https://1drv.ms/f/s!AiOUCEPlhamyhdsPNtnc6Tz_NZm0Tg
Anyhow, why the temp cores are almost equal up to the 50/60C, just at ~90C the temp offset is so significant.
I also applied the generous portion of paste on the right side at my repaste act 3 that i made 1 week back.No change on the temp offset i think.
https://1drv.ms/f/s!AiOUCEPlhamyhdFUOftdH-79rkf_3A
(The file name contains the stress method)

 

BTW, i've checked iunlock's 6820 OC thread and he seems to have nearly 90c on his cores with 4.1 setup on wprime.i cannot find his tests with prime95.practically, my unit is running at the ~89 on cores 0/2 with 4.0 setup on prime95, just core1/3 are cooler...so i am not really sure whether i can expect the core0/2 cooldown at the level of core1/3 once i get this offset issue solved.

 

Hello, Great thread.

I tried some stress tests on my R2 15 6820HK (never re-pasted) and got these results. What you guys think? Seems like quite a big difference between cores

 

That's a huge difference...

 

Whats your highest temp in Witcher 3?

Sent from my E6853 using Tapatalk

 

I decided to return it but honestly, what if the new models have the same issue?

Sent from my E6853 using Tapatalk

 

It depends on if you are talking about AW or MSI, or another vendor.

AW is special in that you can get them to replace the motherboard in house or swap for a new unit, if you got the in home service.

In that case AW replaces or swaps until you get one that works.

There are plenty of people getting good 6820HK units from MSI / elsewhere, a higher percentage than fail, we hope.

So the same goes for other vendors, swap for another one until you get a good one.

A good one is a CPU that has 5c or less core temperature differential when applying the same load on all 4 cores.

Then if the overall temps are too high, you can re-paste, if that's your thing, and drop all 4 cores in temperature.

Everyone should undervolt their Skylake CPU to get better temps overall, so far all BIOS's supply too much voltage by default, running the CPU hotter than it needs to run.

Intel Extreme Tuning Utility or ThrottleStop can easily add a negative voltage offset to the CPU Cores and the Cache.

Start at about -100mV for a Cores voltage offset, and run on that for a while, then bump it down another -10mV to -110mV, and run on that for a while.

Some have reported as high as -200mV on Skylake CPU's.

You can drop the CPU cache a bit too, after you have established the Core negative voltage offset is stable.

Start at -25mv, and work down -5mV at a time checking for stability like the CPU negative voltage offset.

You can wait for a while, and watch for user reports to stabilize too. If you don't want to turn around laptops right now.

You could also buy from a boutique vendor and ask them to test it for you before shipping to you.

Ask them to run tests to check the CPU core temperature offset - tell them you won't accept anything higher than 5c core temperature differential.

Good luck

 

Well, it's Alienware. can I keep retuning the item till I get the one that works?

Sent from my E6853 using Tapatalk

 

Work witih Alienware, explain to them the findings about the "hot cores", how it limits performance - thermal throttling early, and how you need to get a good one that doesn't do that.

If you've already returned yours, you will need to order another one, so you will go through the build cycle again, I assume.

If you've only scheduled a return for refund but haven't done that, check with AW and ask if they will instead send you an in house replacement tech / motherboard. That might be quicker.

Talk to AW about the options, and which is quickest to get you a working unit.

Please come back and let us know how it works out

 

They actually will give me a replacement (new). Currently it's under production.

I really pray the new unit will be in perfect condition (no overheating, no lightbleed, dead pixels etc.)



Sent from my E6853 using Tapatalk

 

If you are otherwise happy with the laptop - good screen, no other defects, then that's why it's good to get a motherboard replacement - or other component replacement - rather than sending back the whole laptop and getting another one.

 

I got to ask, are all AW 15 6820hk owners having the same issue?

Sent from my E6853 using Tapatalk

 

That's why we put up the poll, to find out some kind of percentage of who has what level of CPU core temperature differential.

Please cast a vote with your rating for your CPU, especially if it is good

Currently there the poll shows 30% good, and 5% more on the border.

It will be more accurate if people without a problem vote

 

I just did

Wait, we got other owners who have working NEW AW 15 with 6820HK???? So I guess I'm part of the bad batch??

 

It's unknown what the cause is yet, batches of 6820HK's with a fault that shows up as a percentage of hot core CPU's, or bad cooling hardware, or bad assembly.

The good and bad ones are mixed in together, at least that's the current assumption.

If you got a bad one, you're next roll of the dice will hopefully come up with a good one

 

I just pray the new unit I will is an excellent unit. I will provide updates.

Sent from my E6853 using Tapatalk

 

My A15 R2 with a 6820HK has wild temp differences between cores, but no where close to overheating, once I rerun my tests I will vote in the poll. debating whether to return the whole thing or get it fixed. My A15 R2 is 32GB ram, 6820HK, R9 M395x, 256 nvme ssd, 1 tb hdd 4k screen. Don't think I like the 4k screen, I was planning to just replace the screen with a 1080 for $65 but then I found this article.

 

In my 6820hk all coares are within 1-2ºC of each other, under load or at idle.
Laptop is a Asus G752VS.

 

Whats your temps, brother?

Sent from my E6853 using Tapatalk

 

Hmm, just ran it again and see 84,80,82,75 that is with GPU test AIDA64. Third run 92, 84, 87, 77 also with GPU.

 

Can you try playing intensive games on it?

Sent from my E6853 using Tapatalk

 

Sigh. One last run. no gpu 91, 77, 86, 70 21c swing. Boooooooo.

 

Here are the readings from two separate instances, on different dates ofcourse

http://i.imgur.com/9GHRzhy.jpg

http://i.imgur.com/W7i2RJG.jpg

101C and 102C i have attained.

Regards

 

Your Core temperature differential is under 10c, not great, borderline bad.

But, your overall temps are way to high, stop using it until you can re-paste or return to get another unit, it's not acceptable.

Sorry you got a bad one, come back and let us know how your new unit or repaired unit runs, and please vote in the poll if you haven't already.

Thanks !!

 

Hmscott, my temps are originally posted at

And yes i have asked for a replacement. Will be getting a call in the next 48 hours from dell. I had posted the two imgur links for GTO_PAO11

 

I have a 17 R3 with the 6820. I've read through the posts in this thread. In my case, I have core 0, 1 and 2 running fairly even within 5 degrees of eachother, but my core 3 runs about 10 degrees cooler than all the rest. The one that seems to heat up the most is core 1, but again, usually within 5 degrees of the others. From what I've read across the internetz, this seems to be normal behavior.

 

I'll make sure to update the OP with:

*****PLEASE MAKE SURE THAT YOUR MULTIPLIERS ARE ALL SET THE SAME*****

The stock profiles have the multipliers set unevenly so of course this is going to make core 0 hotter and than the rest.

Make sure to go into bios and set all four multipliers to 36x for example.



::iunlock::

 

Shouldn't all 4 cores work simultaneously in multi threaded bench like Wprime and Cinbench?

 

Could someone explain why those temp spikes during an OCCT stress test on my brand new Alien15R3 with 6820HK Gelid repasted :

The Hot core


The Cold Core :


They seem to be linked to OCCT changing the amount of RAM used :

 

Ok, I got some news for you, friends. I am not sure if they are trying to fib me or what but this is what Alienware Tech Support said:

"We were actually checking the maximum temperatures for this processor and according to Intel and our Engineering team, this new generation of processors can run up to 100 C degrees completely fine."

I don't know what to say.....

 

LOOOL. What with the throttle mess who start at lower temp before thermal shutdown around 100C? What a Joke. + rep You make my day

 

It is funny. Scary at the same time cuz that means tech support will not fix overheating temps then.

Sent from my E6853 using Tapatalk

 

That has totally made Alienware into a tool turd gaming brand IMO... Total fail..

 

Maybe Dell and Intel have collaborated and created new Throttle algorithm for processors in the new Aw machines? You never know, LOL
Most processors work fine up to thermal shutdown? And You don't need to re-paste, because the *new generation* processors work fine up to this temp Oh' well. Finally learned something today also

 

You are still getting a Alienware replacement or RMA though for the Core temperature differential, right?

That's a different problem than even core temps hitting 100c.

At the least you can return the laptop now (like right now) as it doesn't fit your needs for a full refund.

I would do that right away, otherwise AW's stalling tactics will put you outside the easy return / acceptance period, then you will be stuck dealing with AW long term on this problem.

Follow the date's, make sure you effect a return before your acceptance / return period expires.

 

Completely stock 17r4 with a 6820hk and gtx 1070.

Runing an AIDA64 cpu and gpu stress test gave me temps of 86, 76, 81, 75, at the time of the screen shot with the maximum and average temperatures clustered a little closer together.
http://imgur.com/a/Zsd3r

Do you guys think I should keep it or send it back?

 

I'm getting a replacement and it will ship on11/4 but it's weird. When I checked the new order number and the itemized checklist, it does not include the premium warranty and accident warranty. I need to contact them tomorrow about this.

Sent from my E6853 using Tapatalk

 

That's good to hear, so it's not as bad as it sounded - it sounded like AW was denying your replacement / return because the "engineers said 100c temps are ok".

That would have been very bad.

Glad to hear AW are still RMA'ing a replacement for your bad CPU / heatsink / motherboard / laptop

 

I'm not detail informed about what OCCT uses to generate load, but it looks like it's processes started to match cores/thread parameters, and when those processes exit new processes are started.

As the old proccesses exit, new ones are started, and it looks like the timing is such that there is overlap where more processes are running during some time periods then others - maybe it interleaves jobs?

So the load jumps up as new processes start while the old ones are still in memory doing work, raising memory usage and generating more heat than when fewer processes are active.

Those temperature / memory usage spikes are the transition points.

And, that's also why I suggest logging monitoring data to see what's actually going on between the "low", "high", and "average" temperatures. To find such correlations, but also to get an idea of the real sustained load a task, game, or test is generating.

Max/High temp readings don't tell the whole story.

Quite often with benchmarks hitting the CPU hard immediately the laptop fan curve won't kick in right away to take away the heat rapidly enough to keep the CPU from thermally throttling.

After the fan curve catches up and the fans are removing the heat, the Thermal throttling point isn't reached again during a run, unless the run idles for a while and then ramps back up - with the fans ramping down and ramping up again.

That's also why for heavy duty load testing it's best to run with 100% fan's well before and well after the test toallow the cooling system the best shot at keeping heat under control.

 

Alienware haven't implemented fan control app in their main software ACC. Same with forcing fans on max. No keyboard macro for this either... AW owners must rely on 3rd party's software like Hwinfo64 manuel fan control While most other brands models have this already implemented long time ago!!

 

Your image of hwinfo64 didn't include the lines with Thermal Throttling data, but your temps are well below the 92c/93c Thermal Throttling range.

But, I don't know what your usage pattern is, if under heavier load than you documented you could push the temps up higher with the hot cores reaching the thermal throttling zone 10c+ earlier than the 2 "normal" cores.

That wouldn't be good.

So maybe do more testing to see if benchmarks, stress tests, gaming, etc drive hot core temperatures into thermal throttling.

If thermal throttling happens then I would request another unit.

 

And too low headroom for higher ambient temperature and dust as well in the internal hardware, will be a problem sooner or later. Be at the temperature limit when the laptop is brand new...

 

Tweeted Umar from AW. a new Bios to reduce temps is on the works. Let's hope it remedies the situation permanently

Sent from my E6853 using Tapatalk

 

@Vendid

That's also important to take into consideration, degredation of cooling over time.

What might be borderline now could become a problem later.

It's best to accept a new laptop with the optimal function from the start, and not accept a disfunctional CPU.

You only have a short time during the brief acceptance period to test a laptop out to see if you want to keep it, swap for another one, or return if for a refund and get something else now or later.

I wouldn't allow myself to get stuck with something I am already unsure of in the short time I had it, especially not with all the reports of similar problems from many other owners.

I would return it and let AW work out the issues and when that is done then re-order. But, that's me.

You could take a roll of the dice once, to see if AW can swap you a good one, but I would get it agreed that you would still be able to get a full refund if that one also turns out to have problems

 

I ran a combined cpu/gpu stress, so if AIDA did stress both at the same time to 100% to reproduce a worse case scenario, as of now the temps are acceptable. The second graph with the cpu usage/cpu throttling had my throttling at 0%, but the cores only went to x30 multiplier so I'm not sure if there was some throttling.

Things do not look good when you take into consideration the future ambient summer temps. It looks like I'll have to give them a call and see what they will do.

Edit-bad grammar

 

In addition to the excellent advice that hmscott offers, I would also consider just simply returning it, period.

Then, buy a new one (so your return period start new with the replacement notebook).

Otherwise, it could turn into a he said she said kind of scenario and the usual winner will not be the customer, in the end.

Good luck.

 

Thanks @tilleroftheearth - I actually did say said that:

"I would return it and let AW work out the issues and when that is done then re-order. But, that's me."

And, followed up with the alternative if RMA'ing:

"You could take a roll of the dice once, to see if AW can swap you a good one, but I would get it agreed that you would still be able to get a full refund if that one also turns out to have problems"

As far as returning and then immediately re-ordering, that involves a very long delay - the build cycle needs to start from scratch again.

Often people complain they don't get their money right away from the refund. And, they don't have the extra $$$$ to re-order immediately in parallel. So they end up waiting for weeks for the refund, then start the build cycle again.

It's almost Xmas season, so I would imagine the build calendar is going to fill up quickly. It might be after the 1st of the year before some got a new build cycle complete.

If you already have your 1st built laptop, and it's bad, you can return it and get another one quicker than a refund return and reorder, and you have a return period for that replacement too

Always find out the return period details, when you first order, and when you RMA - to make sure you know the exact dates for the return period in all situations.

AW is pretty good about returning for refund, but always establish that as a possibility when buying and RMA'ing so you know your date for a refund return.