[Alienware 17R4 / 15R3] - Disassembly + Repaste Guide + Results

So can I use kryonaut in 0.1 mm space? Is it dangerous?

 

no it isn't

stop being so scared



you can put k5 pro later down the road, they sell on ebay and ship from greece so it takes time to arrive

 

Thank you, very much bro.

 

Hmm i myself dont recommend it to be honest...then at least try to get icd7.

 

Thannk you rinneh. I have ARCTIC thermal 0.5 mm pad. Can I use it there? I do not know how soft is it? Wonder if it can solve the problem.

 

They are pretty soft. When assemblign it squeeze it a bit with your fingers and check for gaps with a light on the other side and look if some light shines through the motherboard and heatsink from the side.

 

Ive repasted, re padded, tried playing with the tension arm. but no matter what i do core 0 still will slowly climb up to the point of throttling depending on the core speed. the rest of the cores stay reasonable with core 3 always being 10 degrees cooler.

not really sure what to try now. debating on getting a hold of another heatsink and seeing if its any different but wont hold my breath haha.

 

If in warranty get a motherboard replacement.
Here’s one shot of the kinda temps I’d get even after repaste. And the rest are now as of today on stock paste


Sent from my iPhone using Tapatalk

 

I finally repasted my laptop with thermal grizzly kryonaut and changed all the pads after correct measurements. But I am still facing the same 1 sec lag issue. I have tried to monitor it by using GPUZ and found the gaps in PerfCap. I cannot understand where the problem is? These are my results after repasting:



Please help

 

Did you put thermal pads on graphics card's VRAM and chokes/VRMs?

 

yes, accept those 0.1 mm tape, because it is not available

 

Funnily enough I’m having this issue.
When it was stock. I had no noticeable throttling. But it was there. I’ve repasted and now it’s much worse. But my temps are waaaay better. Basically in xtu the current limit throttling is fine and same for power limit throttling.
But it’s the thermal throttling that’s lights up. But not often yet. Even when running a game and the xtu stress test at same time. It barely flags up on it.
Yet in games it spikes FPS mad.
But on my last motherboard the current limit throttling was happening constant and power limit throttling would flag up. Yet still the game performance was much better but was just hotter.
So what gives as the scenarios was the same on both boards. Nothing modified. 10c better temps. Worse FPS.


Sent from my iPhone using Tapatalk

 

When does lagging/stuttering start?
Check out this thread, it helped solve the problem on my machine hope it helps you too:

Alienware 17 R4 GTX1080 in game ra...-random-Stutter-&-Freeze.804499/&share_type=t



Sent from my iPhone using Tapatalk

 

View attachment 158180 View attachment 158184 Guys I think I finally found out where the problem was.

The thing is I undervolted my GPU by using MSI Afterburner to improve the cooling of GTX 1080. And this is where the problem starts. I realized, the game we play require much more frequency as compare to idle state. what I was doing is pushing the gpu to work at higher frequency components at lesser provided voltage, which is called undervolting. Unfortunately, this doesn't actually work on most of the laptops including AW 17R4 2017. Where as slight undervolting actually works maintaining the straight line profile. I used GPUZ to observe the PerfCap and how it responds not only for undervolting but also overclocking. Overclocking does not work me still working on it. The screenshot was taken during Crysis 3 game play:

View attachment 158180

Take a look at PerfCap Reason. VRel has no drops to Idle.

The game required 1.04v to run at 1800Mhz. So I forced it at lesser frequency at the same voltage. By doing this laptop would consume slightly more power if it is working in the right region to 1v, and would by slightly undervolted in the left region to 1v.

Also, for the users who are using this machine in hot climate ( ambient temperature > 32 degree Celsius ) must not attempt to overclock their gpu. For me it does not work maybe because my room temperature is 37 degree Celsius.

But my problem is not eliminated completely. Still cannot understand what is Pwr.Vrel. The point where it occurs frames drop suddenly, different from 1 sec lag.

My question is how can I eliminate this full load thing ??

My final findings are:

 

Is the thermal pad thickness pattern for 17 is the same for 15? newbie here I just got a 2nd hand 15 r3 1070 never been opened

 

No Bro, even two 17R4 has different gaps. you can take Iunlock guide as a reference point, but cannot follow blindly all the pad thicknesses he has recommended.

You have to measure gaps of your AW15 yourself. Once padded do not apply paste, use some tweezers to examine those applied pads, if they move then add 0.5 mm thick pads and examine again. It is the only way to rectify the problem, because it is just not possible to observe the gaps with the naked eye.

When you see none of pads are moving with tweezers, then apply the thermal paste.

also keep in mind that adding too much thickness to pads would lift the sink from the memory chips and other ares, so apply the pads with measured and correct thickness. The desired pad with correct would be having an imprint of an IC on which it is placed.

 

Came across this on a random and this video popped up not even browsing computer stuff.



Be very useful in time and money not having to replace if your not one for LM


Sent from my iPhone using Tapatalk

 

Just a shoutout to iunlock for the amazing guide! Dude you need to get paid for this guide!
(FYI I had 11 black screws on step 2).
I did all but the PCH mod(I forgot to buy the stuff, or maybe I couldn’t wait).
Amazing temp drops up to 20+! In some cases I had differences of 30C playing Arma3.
My passmark score increased as well!
When I replaced the four little square thermal pads the stock ones still had those plastic peel off protectors on them
I’ll upload pics when I get home.
I can overclock and feel good about temperatures again!

 

thank, is there anyone have done a pattern for AW15?

 

Just tell me How are you goin to use it on a laptop cpu die 1/4th the size of the one in desktop?

Another question is gpu has capacitor all around the die, and this one is conductive. If we assume to cut it, for the dimensions as that of the gpu die, the value of thermal conductivity would fall down to the approximate value of the ordinary paste.

So, what is the advantage of risking the 3000 dollars laptop just for the sake of some innovation ?

 

In the video it does state you can cut it down. If they saying this then surely there is no loss in performance.
Plus it’s reusable. Was apparently tested 50 with the same pad and no change.

Maybe in time they’ll cater specifically for laptop market where they’ll have it cut to size.


Sent from my iPhone using Tapatalk

 

But bro the problem is not the cpu cooling. The problem is the Power delivery circuitry cooling i.e. the VRM section of GDDR5X.

I am confused, every body here is saying that not to worry about the Chokes ( Inductors ), the main problem is to provide cooling to tiny MOSFETs just above the inductors. I have recently study a bit about the PCB design of NVIDIA GPUS and found out that those inductors which we are neglecting are a part of VRM circuit.

So according to that, the full VRM section ( FETS, Inductors and drives ) of the GDDR5X must be properly cooled for " Load drop to zero solution " problem.

Correct me if I am wrong

 

Not going to correct you there. Have seen people put paste on those chokes. So it has a big impact to help cooling.
Never seen this used on gpu tests but in terms of the cpu and differential core temps. It be helpful to ensure evenness besides the heat seat having its own issues.


Sent from my iPhone using Tapatalk

 

Im scared... I am Mostly scared.... Mostly....

 

Take a look at this guys.

This profile is a miracle.



This is a screen shot taken during GTA V game play. The main thing to focus is the power consumption of GTX 1080. It is a good practice to use lesser voltage to perform any task in electronics provided that other parameters remains same.

So by doing this we can achieve 1800MHz approx. at 0.86v resulting into the power consumption of 121.59 Watts. We can go at upper frrequencies by finding the sweet spot of voltage/frequency curve which would result in more power consumption. One should keep in mind the closer the value of power to 180 Watts, the greater the chances of shuttering.

It is my opinion that the reason of shuttering even after re-pasting and re-pading is an inefficient design of VRM section of laptop. I do not know which quality FETs they use to deliver power to GPU and GDDDR5X memory chips. I also do not know the no. of phases this design is based. Are there two lower MOSFETs in memory VRM or a single one? these are all the questions that I think may be able to understand the inefficient design of VRM. And due to these questions, I decided to to reduce the power consumption of GPU.

Also, I think that increasing just 100MHz at the cost of 50 Watts Power consumption is not a wise idea.

Now, I understand what the new BIOS did is to remove the voltage barrier, limiting the value of 150 Watts power consumption. I do not know who to blame, but for the laptops where we have limited PCB space and lesser efficient GPU VRM design as compared to the desktop GPUs where there is a lot of space, we should not use Unlocked BIOS.

I may be wrong but this is what I think.

 

I would like to thank iunlock for the solution on this thread. I tried first kryonaut ...was way better then original cooling solution. After 6 months I took the risk and after I manage to get all the requirements for upgrading to LM I got to it.
So far I cant say how much difference its there in our grasp. Or better to say how bad the original paste was... not to mention the tripod cpu heatsink design. So far I am using at 3.7 Ghz cpu at 35-38 idle with internet surfing and wireless. After testing with OCCT about 50 min I got 67 on cpu and Superposition Benchmark - 4k optimized I got 70. Good enough, but then...I've gone to PUBG and set everything on high and start gaming. All this time hwinfo opened and monitoring temps, fans speeds, power usage. The highest temps were 72 on cpu and 70 on gpu. Another thing... the m.2 ssd 960 pro is getting so hot on this game. More than that, I cover the PCH with a raspberry pi cooper heatsink, but I didnt use LM on it or build a socket for it. The heatsink has some self-adhesive paste but not so good cooling, cause I still get to 80-85.
If someone else wants to use this as well... be careful on the height -its about 5.5 mm so you need to use a sand paper and make it to 5 mm or lower. It fits exactly.
As a final word I would like to give a big shout out to all the others who have shown support with the AW17R4 project.
Thank you guys.


p.s. I live in Romania, so...anyone living around and interested in repaste and repadding his Beast can pm. I can buy all the tools from here or you can order them from the links I can provide.

 

i just used that thermal pad posted above. it gives me the exact same temperature difference problem still core 0 will shoot up 25c higher than the rest.
Im hoping its just a really messed up heatsink that isnt level. I have tracked down one online for cheap that is one of the 'better' versions made so
i guess we will see what happens next. I really dont want to have to deal with warranty a 3rd time with this laptop. (already had 2 motherboard replacements)

but if i have to then i will have to.

 

Sounds like a mismatch between motherboard revision and heatsink. The power delivery section of your cpu seems to push up your heatsink if the differences are that high.

 

Hi i repaste my alienware 17 ( i7 7820hk and gtx 1080) with grizzly conductonaut . When i play GTA V cpu reaches 88 degrees and GPU 70 degrees (°C). If i stress the cpu to 100% it reaches 65 degrees. Is it normal?

 

Yes..its normal...cause when you test stress the cpu the gpu is cooler then the cpu and doesn't heat...and they use the same cooling heatsink. When you play GTA...PUBG or something demanding on both... the heatsink is not cold on gpu side... they both heat up the heatsink. Same goes either way. Stress cpu or gpu they are not going to pass a very high temp. When stressed together they heat up. Try undervolt the cpu and let it at 3.5-3.6 ..it would stay cooler then 80. For the undervolt, underclock, overclock... check forums... dont do it by guessing.
Good luck.

 

Thank you for the great guide! But I have a problem to disassemble the head sink. After what I have read I should use PH1 bit. But I`m not able to unscrew the headsink. Seems like this require some other bits? Any suggestion?
This is how the screw look like:

https://ibb.co/hKpevo



Also seems like I damaged one of the screws with the PH1 bit:


https://ibb.co/eLSGao

 

Don't hate me for not reading through 334 pages.

I'm going to repaste my laptop, I'm just trying to figure out longer term experience with LM. I'm not concerned about damaging any components as I have extensive experience watercooling and working on pc's, but I'm extremely turned off by the idea of needing to perform periodic maintenance for the LM.

Hoping someone can either provide some feedback/personal experience, or can point me to a better thread with people's results?

So far, about half of what I read says it's good to go with no issues ever after installing, and half says it'll need to be redone every so often. I understand the difference is primarily if the heatsink is a copper block as the LM reacts with that. I believe my heatsink is nickle so maybe this is a non issue than?

But I've also read people saying that the heat cycling and hitting high temps with the LM can force it out, therefore requiring maintenance. I'm not doing this to an alienware (asus GM501 instead), and I do expect, even with LM, that my temps will hit over 80C so that may be an issue.

Any guidance/feedback would be appreciated!

 

Question: The liquid metal seems to be absorb into the heatsink after about 5 months which makes it looks like it's dried out which make the temperature bad again. Should I reapply without removing the old liquid metal?

 

Mine looks like it's dried out after about 5 months. I guess you need to reapply it when that happens but I'm not sure if you should clean the old liquid metal or just apply another layer on top of it.

 

I think the best is to remove old LM (dust particles and chemical reaction impurities, dry LM), but leave the stain (only discoloration) on heat-sink. If you put fresh LM, the heat-sink does not/cannot absorb no more LM over time, because Copper and LM are already bonded in the stain.

 

If you replied to my messages on battle.net or even asked me, I would have been able to tell you.
NEVER reapply on top of dried out LM. Contact will be nonexistent, instant 100C.
You need 100% or 91% alcohol, and 3000 grit "polishing" sandpaper, that will remove all of the old alcohol without wearing down the heatsink.

Part of the LM will always harden. To keep the main part liquid, you must:
1) have good contact pressure.
2) use enough LM to compensate for the battery effect (meaning you have to use more than the "super thin" layer others suggested). Foam dams will help protect from any excess runoff.
3) rub VERY vigorously a coating of LM On the heatsink surface initially, using something lint free, and work it in hard and fast. This starts the battery effect directly on that layer rather than the CPU.
The conductonaut swabs will not work well for this. You will have to wipe so hard, fabric might rub off and any fabric or fiber on the heatsink isn't good.
I recommend something like this for spreading and rubbing in LM with no fabrics tearing off or fibers:

https://www.amazon.com/gp/product/B00Q3BG6Y4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&th=1

 

Hey dude, sorry for that, been busy with College and I just finished with exams. Oh and I don't have that battlenet accout anymore (gave it to my friend cause he wants my hearthstone account) so you won't be able to reach me on that. Thanks for the answer. But isn't the foam will make the contact between the cpu and the heatsink even worse? I mean unless I could find somekind of supper thin foam.

 

In Asia now with ambient temps of 25c/26c in my current room. Getting a slightly higher core differential than back in Europe. Which I want to rectify. It annoys me to no end now

But my core temperature distribution is different than the usual "tripod" pattern.

My core temperature pattern looks like this when stressing the CPU

core 1 - 70c
core 2 - 72c
core 3 - 67c
core 4 - 67c

Is it safe to assume that the GPU side of the heatsink is causing the uneven pressure in my case?

 

Check C0% and your background apps. Most apps tend to push 1 or 2 cores which might explain uneven core temps sort of issue.

 

Thanks for tbe tip. But it is in prime95

 

Never tried it. I use my own crappy app to find out best undervolts and stress test. Most of the time with AVX workloads even I have 1-2deg core difference on actual cores and 10C on virtual cores.

 

These are the cores, although its numbered from 0-3.
Check the numbers (in your case 0 is 1) and see where the problem is.

 

Thanks, I thought it was in this order but you confirmed it. Must be the GPU thermals pads next to it then pushing that side slightly up. All core temps are good and close to eachother except the second core (core labeled 1 in the pic).

Sucks that the thermalpads arent very compressible on the heatsink. Gonna replace them with Gelid pads.

 

repasted my R5 with i9 8950HK and GTX1080 yesterday with LM and new pads. Thanks @iunlock for this awesome guide

cinebench 4.5ghz


Far Cry 5 15min warm up @4.3ghz then 15min 4.5ghz test

 

Hmm thats good to see actually.....gives me faith in this design. So with the right paste and thermal pads this machine can do exceptionally well. Even though a lot of people have shown good results, some people claim it is impossible.

 

4.5 ghz @ 1.1v VID ? At full load?

I am calling Shenanigans on that one.

 

that is with undervolt

 

Hi all,

I previously did re-paste and re-pad, but because my heatsink is badly fit, 1.5mm or more thermal pads are required in many places so even though I use all Fujipoly 17 W/mK pad (a real pain in my wallet) it is still not efficient enough. Half a year and they dried out because of extreme temp (it is almost always 30+C in my country).

Now I am thinking of SOLDERING COPPER PAD to any places where a thermal pad of 1mm or above are required (1mm copper pad for 1.5mm gaps, for example). After this process using ONLY 0.5mm THERMAL PADs is sufficient to fill those gaps.

I did have a research through different kinds of solder pastes and decided that Sn63Pb37 is the paste to go. It has melting point of 183C and is one of the most thermal conductivity.

The only thing that worries me is WHETHER THIS SOLDER PASTE CAN STICK TO THE "GREY" ALLOY OF THE ALIENWARE'S HEATSINK OR NOT.

So, could you please give me advice on this?
Thanks so much.

Edit: Sorry but I will mention you here @Papusan @iunlock

 

Most likely tricksing with vcore. Probably same nasty as MSI. See running +55mv higher voltage but 6% lower power consumption for equal clocks. You find the same for the 2017 Aw models as well.