How to FIX AW17R4/15R3/13R3 CPU Core Temperature Differential Issue

Purchased "new" from the outlet last week 6700HQ 13r3. Core split was about 5-7c* prior and after the repaste. Temps were not great but within spec at time of purchase; Sunon HSF.

 

After Months of 6-8°C temps differential between core 0 and 3 @ stock frequency. I managed to reduce it to 3°C.
I removed the 0.1mm pads and replaced them with dots of thermal paste for each VRM/Chokes.
The Core 0 temps decreased, but the Core 3's increased too. Could be explained by the pressure moving on the curved CPU surface.

I wonder if LM could resolve this issue without spilling out. Or if i need to lap the heatsink, not flat but curved to match the CPU curve.

 

It depends on your apps too, not every app will load balance itself across multiple cores. Can you provide a screen shot of HWINFO?

 

8mn OCCT 8 Threads to check quickly the diff:


But now, has soon as i OC to 4.0GHz, the temps skyrockets higher than before! And i think this is because the heat dissipated at the center comes from the Ring Interconnect and only half of the Cores :


Before it was coming from the Cores at the bottom:


Look at where the scratches are on my previous CPU with 7°C diff between core 0 and 3:

 

why not try to even out the temperature difference with a thin copper shim?

 

This could work, but i'm not fond of the cpu-tim-copper-tim-heatsink sandwich. It isnt optimal for heat transfer.
LM looks the most optimal way for my CPU, but i would have prefered to fix this issue without it.

 

Or try a new TRIPOD.

 

I've seen a few listings for these heatsinks which have images of Sunon fans and other side stamped CCI, seems very strange.

I'm not implying anything....but with reports from people having noise issues with Sunon fans and misalignments with CCI heatsinks, could this possibly be cast off of defective parts put together by people who received replacement heatsinks from Dell ?

 

I've seen a few mentions of the "skylake bezel and webcam" in this thread. Did Dell actually change the bezel and webcam on the Kaby Lake laptops?

 

updated thread #1 with new info on different heatsink assemblies. feel free to share

 

Dell replaced my 17R4 because of several bad pixels on the screen. This week I received my replacement which has a build date of 9/28/17.

I immediately opened it up to install my ram and SSDs and became sad when I saw that I had a CCI heatsink (my previous was sunon). I was worried that maybe they sent me a refurb because I thought they stopped using CCI?

Well my worries were for nothing because the temps are GREAT!

After 1 hour of PUBG I had CPU core temps between 62-64 and GPU also only hit 64 max. This was without a repaste, just propped up on a laptop stand for better airflow.

My heatsink is a DP/N 0FRPY8 rev A00 and says CCI BL-2 on the GPU block.

 

AWESOME! thanks for the feedback. i'll update the main page with your info! looks like 0FRPY8 is the latest heatsink revision made by CCI.

Have u tried OC to 4GHz+ with 100W PL1&2 and check the temperature when under OCCT stress test?

 

No, not yet. Once I got it set up I just started gaming. Once I'm happy that everything is stable I'll try an OC but this thing is fast enough in games that I probably won't need the OC for everyday use.

Here's some pics of the heatsink. You can see why I thought it was a refurb because of how the heat pipes look rough on the left side and there's marker on the frame.

 

This is the Heatsink i received with my laptop. it also gave me temperature differentials but they where between 5 and 10 degrees. Bad but not dangerous, i already had problems with the temperature spikes with this heatsink. The differentials just got worse when i received my new heatsink. I think before i'm going to repad i'l try to add a washer to the screw for more pressure on the cpu die. I also saw a post once on reddit where someone added a copper shim of 1mm to even out the pressure. he then proceeded to repaste (with kryonaught) wich severly got his temperatures down (in combination with the shim) i'm going to look if i can create an user friendly DIY for the pressure by adding copper shims to the heatsink and then repad the shim. so you'll get somethink like Heatsink -> copper shim -> thermal pad -> finfet chips. And on the cpu it would be Heatsink -> copper shim -> thermal paste -> cpu.

 

thanks for the picture. looks like your old heatsink is 04RFW1 Sunon made but im not sure what revision was that (S-J-1 or S-J-2 perhaps?) even though DP/N stated A01. i can save you some time.
copper shim method, dont work, well not as well as we both hoped it to be. i have tried it. at over 4GHz OC, the thin copper shim sandwich method still gives more than 10C temperature differential.

 

Thank you for your reply, hmm there should be a method to even out the pressure (without using LM) i travel everyday with my laptop so LM is not an option (scared for leaks etc).

The heatsink is an S-J-1, ordered my laptop 24 juli got it on 8 august. (today is our 2 month anniversary lol) and since day one i've been testing, benching and undervolting (while having a oc'ed speed).

i don't have any problems with me cpu hitting it boost speed (max oc is 4.4 ghz on all 4 cores) but as soon as it goes beyond it's powerlimit (it surps around 55/60 watts while gaming, benchmarks around 65 watts) so it's not even using it's full powerlimit and it already starts to throttle. and the funniest part is when it throttles, it goes back to 3.9ghz to shoot back up to 4.2 (on all 4 cores) and it continues.

the games where i have the most problems are Flight sim world and Flight simulator X. both are simulators that are unoptimized so maybe that could play a part, but there are people on youtube with 6700 and 7700k playing without problems. and since the 7820hk is roughly on par with the 7700k on the same speeds it strange that my cpu behaves like that...

i turned on speedshift i throttlestop and configured to my needs, and it has helped a little but not much..

 

Shims are a waste of time for this and will complicate pad thicknesses as well by making all the gaps larger

Sent from my SM-G935F using Tapatalk

 

i previously got a 04RFW1 Sunon S-J-1 too. rubbish heatsink, almost 20C difference out of the box. even with a repaste i cant get it to be less than 15C without some serious modification.

try reduce the OC to 4.1, and undervolt -100mV. trust me, you will be a lot more productive.

 

i tried and sadly my cpu doesn't work like that my temps stayed almost the same with thermal throttle on core #2. i know the cooling system is capable enough to cool my cpu at 4.2 or 4.3 on all cores.

core 1 and 3 are cooled propperly so it's evident that heatsink is warped. it almost does not matter at what speeds i run, core 0 and 2 throttle very fast ! I also lift the back everytime and it only happens when playing cpu intensive games (or let windows defender scan my ssd's) as soon as it starts to slurp power, the temperature shoots up. I think a repad is my safe way to solve this problem.. else i have to close a pact with the devil, and repaste with LM but i move very often with my laptop. I go to school and work (chip tuning cars and motorcycles) so i'm scared for leaks (even though i'll be using tape for safety meassure)

thanks for you suggestion though, i think i'l just see how 4.1 performs and see how much i can undervolt it. Very often my undervolt passes in OCCT but crashes when using Visual studio or when i let it idle/go to sleep.

 

I am AW15R3 owner. My laptop was shipped on Sep 28, it came with a CCI heatsink.

There is also a core temps differences which is around 10 C after use of thermal grizzly hydronaut.

However, I am seeing a person did a AW15 1080 max-q test on Youtube. The test machine has no core temps differences. The heatsink used on 1080 maxq is a little bit different than 1070. You can check out the videos.



The left one is 1080 maxq version.



I am thinking about the heatsink on 1080maxq machine does not problems.

 

I also noticed that there is a hot spot above the letter "E" at the "ALIENWARE" logo under screen. Is it usual or unusual?

 

Max-Q has nothing to do with CPU core temp differentials. Some builds out of AW are ok, few but some.

It's a matter of tightening the 3 screws lightly going in round robin to get even pressure from all three pressure points. And, the right thermal pad thickness needs to be used in a number of areas to make sure things fit evenly.

The AW fixes are many to attain even CPU core temps and low temps, GPU is usually fine. This thread is good, and there are others.

Check out post #1 in each of these threads, started independently over time, but supporting and using what is learned in each to arrive at an overall solution.

[Alienware 17R4 / 15R3] - Disassembly + Repaste Guide + Results
http://forum.notebookreview.com/thr...5r3-disassembly-repaste-guide-results.797373/

How to FIX AW17R4/15R3/13R3 CPU Core Temperature Differential Issue
http://forum.notebookreview.com/thr...u-core-temperature-differential-issue.805062/

Warning: Some i7-6820HKs and i7-6700HQ have Uneven Core Temps due to Uneven Heatsink
http://forum.notebookreview.com/thr...ven-core-temps-due-to-uneven-heatsink.797477/

 

You are right. My point is that the new heatsink used for 1080 can be the latest version for AW15.

 

Does it make a big difference? I thought the GPU temps were already similar to other laptops with 1080's, or lower.

Are the new GPU heatsinks in the AW 17r4 1080 110w better or worse?

The Max-Q GPU put's out a lot less thermal load, so it's going to run cooler unless you manage to unlock it and increase the power limits, then it could be much worse.

The Max-Q 1080 110w GPU is not going to OC as far as the full 1080 OC 200w, and that full 1080 heatsink / cooling has been enough to OC a 200w+ 1080 GPU.

The current AW 17r4 full 1070 125w comes pre-OC'd to perform as well as the Max-Q 1080 out of the box for $400 less.

The whole Max-Q scam is a confusing waste of time.

 

I am not talking about the max-q 1080's performance or overclocking. I am saying the max-q 1080 used a slightly different heatsink than 1070 machine. I am wondering if this heatsink can improve the core temps differential issue. Because AW17 has different heatsink revisions, the different maxq 1080's heatsink should be the latest revision for AW15.

Do you understand? I am not talking about its performance.

 

Yeah, you weren't / aren't being specific enough

Are you saying the CPU portion of the heatsink is different with the Max-Q 1080 model?

And, if the Max-Q 1080 model has a different GPU heatsink section, it makes no difference with CPU core temp differential.

Let's drop it for now, the sacrifice in getting a Max-Q 1080 instead of a full 1070 / 1080 wouldn't be worth easier CPU heatsink adjustments, not at all. Yuck.

The regular heatsinks can be affixed with even core temp distribution out of the factory if they would take the time to put on the right thermal pads, and check the clearance as they screw down the heatsink, as in the video, that's what makes the difference.

 

Which video are you talking about? Can you post it? Thanks

 

I'm refering to the video's you posted

The first one has them pulling the heatsink assembly and trying to get it on right a couple of times, checking the fitment as they went, turing the screws in order, that's the way to do it.

 

Are all the heatsinks on AW 15 R3 and AW 17 R4 compatible ?
I'm assuming the GPU 1060,1070 and 1080 doesn't make any difference.......but I've seen 65w, 85w, 125w and 138w heatsinks, I'm totally confused.

So to be more than specific, if they're all compatible, the One heatsink to get is...........{Name, Parts Number , Revision} please.

 

Shifted for RAM clearance, due to lack of space on Motherboard they decided one arm #1 was sufficient.

Maybe we're all seeing the issue with a wrong perspective, maybe it is NOT the single retention arm #1 being too weak but it IS the two fixed arms #2 and #3 that are too low (as someone had posted), in precision engineering, 1mm too low is 1mm too much.

I have reasonable temperature but still unreasonable differentials...better for lapping the heatsink copper plate bald patches but not prepared to bend the heatsink or lap the copper plate any further.
Found some 6mm washers from an old IBM Model M metal Plate rivet mod. To raise the fixed arms #2 and #3 by 2mm by placing a washer between the pole and the fixed arms (for testing).
The washers at 4mm thickness may be too great a gap to raise therefore I'm going to lap it down to 2mm before insertion (when I have time). If necessary, lapping it down again to 1mm if the gap is found to be too great.

 

You need to repad part by part on that VRM (just above the CPU, I'm unsure what's its called), it should look like a step more or less.
Its good to see there is LCD display with accurate measurement on a vernier calipers.

 

My 17 R4 is from April 2017 (7820HK, GTX1080)

Original heatsink was a CCI, can't remember which one
Overheated, couldn't fix it with repad and repaste, still freezing

In June I got a replacement motherboard and heatsink, this time a Sunon S-J-2, feels a bit heavier than the CCI but I didn't check.
Stock pads and some kryo didn't freeze at all, even on OC vBIOS
Core dif was about 10C. Repaded with arctic pads and new kryo. Temps aren't great, reaching 85C while gaming for 1+ hours. But everything works fine. Core dif was at 4-8C

Switched to Conductonaut LM, bent the CPU arm a bit, checked every pad again (BTW, with the S-J-2 many of the pads don't match iunlocks photos in his guide. On the CPU chokes there's a 0.1mm pad that needs like a 1 or 1.5mm to start making contact).

No more core dif issues! All cores are only 0-2C in difference

Temps were pretty good now but I still figured I'd help it out a bit more. Summer is pretty hot where I live so I undervolted the CPU ~90mv and installed the old vBIOS (did some gaming tests and didn't notice any difference, didn't benchmark though).

Now the system runs great! CPU doesn't reach 80C, many games dont even make it sweat at ~75C like CS GO, Overwatch....
GPU is usually like 10C bellow the CPUs temperature. And I've seen my PCH reach ~83C, still without issues. (I don't have a heatsink on it).

Maybe next I'll follow OwnOrDisown's video about undervolting the GTX 1080 to get even better temps and performance.

It took a while. Too f*cking long! Never have I had this much work just to make a laptop work properly.
But when it's tuned the 17 R4 is awesome. I wouldn't trade it for anything else right now.

 

let us know the result of your investigation

 

Sanded and polished the two washers to 0.2mm thickness last night, doesn't need to be polished that's just my OCD playing up.
The outer diameter of the washers are 6mm and the inner bolt hole is for M2.5 bolts. Hope it fits as I'd not taken any measurements when I last repasted.

Unfortunately I'll be very busy for a while and need to use the laptop, as the temperatures are okay I won't be taking it apart just yet.
Hopefully, I'll sort out my core differential issue when next I repaste.

Credit where credit is due to leeloyd, I'm only trying to prove him correct by lifting up the fixed points #2 and #3. Not wishing to bend these arms in case it creates further issues elsewhere.
http://forum.notebookreview.com/media/alienware-15r3-17ra-bad-heatsink-design.1690/
_________________________________________
Will be most interested to know how those Taobao GTX1080 138W heatsinks perform (fan noise) and if it's compatible with an AW15 R3 machine, so far no one has purchase one as yet.
http://bit.ly/2igLaYT

 

New 15r3 owner here, maybe temporarily... but anyway, I noticed on my 15r3 that there is a larger gap between the heatsink and MB at number 4/6 (bottom of the GPU). I tried taking out thermal pads and putting them in to bring it level, but the gap is still there and I would think it might have some effect on the CPU heatsink displacement. It also tells me the whole thing is basically warped. If you get a chance try and place your motherboard on the corner of a table so that the fans do not support the board. You can lay your heatsink down on top of where the screws go in and it can give you and idea on the condition of your heatsink.

Both of my repaste tries, using kryonaut, yielded about the same CPU results (83-86c), using different techniques (line/spread) and -.125 undervolt. I just can't think of anything else to do, except get a refund.

 

I’ll try this later tonight. My main concern is having at least 10°C after the initial repaste/repad using Kryonaut and Fujipoly.

 

Bad news, 0.2mm had lifted the #2 & #3 far too high that the TIM had no contact at all. Less than 10 seconds and throttled.
It takes too long to sand down these washers as it's steel so I sanded the copper plate down a little more then repasted. Maybe in three Months time, I'll retry with 0.1mm thickness.
The result of sanding the copper plate was quite good, brought down the temperature by a few more degrees but differentials are still niggling...


Does anyone know what was addressed by the recent BIOS updates ?
It seems to run less with less noise, maybe they lowered the fan speeds ?

 

Disable Built in Monitor of OCCT, use Throttle Stop or HWINFO 64. OCCT HWMonitor seems a bit off since last releases. Your temps look good.

 

Just repasted the new 6820/1070 with LM. Temps looks great, except for core 1, which is showing 25c higher temps than the other cores. The 1070 doesn't heat up at all. Does anyone have a diagram showing the motherboard orientation and cpu, and which core is 0-3, so I can adjust the pressure to the cooling solution... (shim or bend tripod, etc)? Thanks in advance.

 

Post HWINFO64 sensors screenshot.

 

The clocks run 3.0GHz, bruh. Of course max temp will be lower the lower clocks sink due throttling

 

I don't know why upon running OCCT, CPU esp. locked ones just stay at 2.8GHz or 3GHz. Disabling that sensors of OCCT does the trick.

 

Clocks as you can see, increased to normal level after the test was stopped/cancelled. And 7820hk is unlocked.

 

I think OCCT is black-listed by Intel. On AIDA64 it works 100% of the time at turbo speeds.

 

Then, why should OCCT suddenly stop working correctly with Alienware’s now? Have worked just fine before. Nothing has changed as far as I know.

 

Not only AW but all PCs with HQ chips. I noted it on my cousin's lappie from HP.

 

Mind you... The screenie above was with HK chips running

 

Yeah I know.

 

please delete

 

Here ya go..... Big undervolt....