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

22c ambient. I am in DC/Maryland


Sent from my iPhone using Tapatalk

 

Can you clarify what test you used? Is the system elevated? Did you unlock the cores of Pl1/Pl2 at stock clocks? Lots of factors that play a role in temps and trying to manage the information coming to the group.

As someone relatively new to the 17r4 repad repaste in the beginning it was hard to determine which temps where made with similar ways I was taxing my system vs which were not. For the sake of other new users we should clarify as a group.




Sent from my iPhone using Tapatalk

 

I ran OCCT and the CPU Stress Test from the SupportAssist program. The rear is elevated as I don't see the purpose of leaving the system flat since there is hardly any room for air to flow.

I have not unlocked the cores of Pl1/Pl2. Could you point me in the direction on how to do that? I've downloaded ThrottleStop since I've read that XTU may cause issues with undervolting. I haven't used ThrottleStop since version 5, so I'm a bit out of practice.

 

When using LM, you really should get free insurance by using this type of exact foam, by cutting and wrapping a layer in the shape of the silicon, around it, to act as a dam to prevent any LM from escaping anywhere:

https://www.amazon.com/EcoBox-Polyu...4366948&sr=8-5&keywords=foam+packing+material

This is the type of foam me and @Mr. Fox have, however we got it from spare packing materials from other items. This stuff should be available at any hardware or crafts/hobby store, but the important thing is to make sure it can be compressed to mere micrometers. You don't want any resistance from the foam to interfere with heatsink pressure, so it has to be as squishy as possible. Dense stuff that can't be compressed all the way down is a no-no.

This is in addition to the normal electrical tape or nail polish people use to paint over the little resistors by the silicon that can be shorted out and destroyed by conductive balls of doom. The dam isn't for those--it's to protect the rest of the laptop.

LM is dangerous if you don't take proper precautions, and a godsend if you do take proper precautions.

 

It's not my first time using LM. I had some On my R2 back in the day.

I am a bit concerned as I'm running into some BSOD (Critical Process Died) while gaming. So far I've only noticed it happening when I either OC the GPU or CPU. I was pretty careful with the LM and made sure to get a nice even coat without spilling any. My temps seem just fine as they don't get past 75 when gaming.

I may need to open up my system again and check it over. Unfortunately I'll be out of town for the next couple days, so it'll have to wait until I get home.

 

Just an update. I am ending up selling my 17 R2 since it just has to many problems.
Bought an ASUS ROG G751J instead, which has 980m and gsync ;3

 

If you have spilled LM, you would most likely have had a dead machine now, not some BSOD. Look other places. Download Who Crashed and take a look into the error log. Google the results.

 

I agree to that. I've read of horror stories with LM. I'm thinking it may be the NVIDIA driver as it didn't seem to install properly. I may be home earlier than expected today. I'll try and put some work into it tonight.

 

I'm noticing today that my CPU5 & CPU6 are about 10 degrees higher than the rest. Right now I've got my BIOS set to lvl 3 OC and I'm playing Secret World Legends. All other temps are sitting between 62-67 but CPU5 & 6 are between 76-79.

All temps are bouncing around slightly, but 5-6 are making the biggest jumps. Occasionally from 65-81. Any idea why this is?

Edit: after about 20 min I got the BSOD again (Critical Process Died). I uninstalled and reinstalled the NVIDIA drivers last night and ran CCleaner to clean out the registry. I did uninstall XTU the other day, but don't remember if I was getting this BSOD before or after uninstalling.

 

Apparently my system isn't saving the error logs. When I run who crashed it says no valid crash dumps have been found.

 

If you have undervolted your cpu, reduce a bit on your undervolt. If Oc'd, reduce clock speed and test again. Test also with default clocks (with and without undervolt). Too low voltage can mean freezing aka light bsod (no error log will be created).

 

I had it undervolted using XTU, but reverted it back to stock voltage before uninstalling it (at least I hope I did it correctly). What is the stock voltage for a 7820HK?

 

I don't have BGA. But expect at least 1.200v for max clocks with load.

 

I'm still relearning how to use ThrottleStop and when I check the voltage under the FIVR tab it was set to .0996. I reset the voltage and will see if that helps with my BSOD. I am noticing that my voltage is reading 1.3+, but that is with the lvl 3 OC in the BIOS. I'll put it back down to stock clocks and see what it reads there as well.

 

Here we go!! Different voltage level for load vs. idle clocks.

 

Voltage still looks to be about the same without a OC. Still getting the BSOD. Looking through Event Viewer I'm seeing multiple warnings that state "The speed of processor X in group 0 is being limited by system firmware. The processor has been in this reduced performance state for 71 seconds since the last report." Gonna do a little research on this to see if there's any possible fix. I may be heading towards a full OS install. If that don't fix it, back to Dell for their diagnostics. The thing is I can run benchmarks just fine, it's when I game that the issues arise.

 

Updated my IRST driver just for S&Gs. Opened up Secret World and started playing. Game was running fine with CPU set at lvl 3 OC. Then the game freezes and closes out. Then Windows freezes and goes to a black screen. After a few min, restarted on its own and getting a Boot failure on device in SupportAssist. Could the issue possibly be my NVME M.2 SSD overheating? I've heard they run the risk of running hot, but it's very hard to get them to high enough temps to cause issues. Plus my games are all on a deprecate drive.

 

Run ATTO Disk Benchmark 4-5 times in a row. Monitor the ssd temp with Hwinfo and post final results.

Run Aida64 as the Guide. Put up Hwinfo as the pict and follow up the hardware's behavior.

Spoiler: Hwinfo setup

 

I think I may have found the issue. My SSD was in the slot marked 2. I moved it to slot 3 and have had no issues while gaming. Perhaps slot 2 is faulty and I'll need a MB replacement or since slot 2 is closer to the GPU, it was hitting higher temps. For now I'll leave the system as is. On Thursday I should be getting my heatsink for my NVME drive. I'll move it back to slot 2 and see if it'll help with anything.

I am hoping there is nothing wrong with the MB as I'll be deploying soon and don't really have time to wait for the system to get repaired.

 

Hello to everyone, I've been following you for some time, especially iunlock that I follow a lot. Sorry for my English but I write with the translator, because I'm Italian.
I have an AW r3 15, 6820hk + 1070. Soon I called Dell Support for the problem of overheating. Not happy with the treatment by Dell, I made him repose, first with the "kryonaut" and then, as nothing changed, I put him the conductonaut.

the temperatures are as follows:

(overclock set to 40 by bios or throttlestop, and without the "performance" function for fans + undervolt with throttlestop at -50mv cpu core and -30mv cpu cache)

Idle 49 to 59 degrees, depends on how long it is on.

After gaming a bit, I get out of the game and I immediately score the XTU benchmark, it also reaches 89-90 degrees, with a core difference of 6-7 degrees maximum.

The gpu with the heat pasta was cooler, now it reaches 78-79 degrees in play (with a bit of overclock).


I am referring to you, more experienced than me: how can I do it? I have the sunon as a heatsink. Do I have to buy another one? I disassembled it several times believing that I had the problem with the screws, but I never solved anything.

how does iunlock have those temperatures? I swear I would pay gold to fix it all. Unfortunately, I can not contact Dell Support anymore since I have run out of warranty with "repaste".

I look for an answer, thank you.

 

Lift the rear of the laptop by an inch using two even sized bottle caps or by using a ergonomic stand.

 

What heatsink for the nvme drive. I would like to mod mine as well.


Sent from my iPhone using Tapatalk

 

I'm going to try placing these on my NVME thats been getting hot.
https://www.amazon.com/gp/product/B01GE7Q060/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1

Waiting on my intel 8265 wireless that should be in the mail today before I open her up so I'll report back later.

 

NVMe ssd Heatsink Mod You will find more in the sub-forums or on the web.

 

I just threw on 4 15x15 1.5mm copper shims and arctic silver pads. Max temps on my 512 toshiba drive don't exceed 68c now... used to hit 80+.

No cutting or sanding of metals required.

 

Installed my new intel 8265 card and bluetooth latency issues with my DS4 have been completely fixed.

On the topic of the heatsinks for the NVME, they were to tall so I didn't place them on yet. Think I'm going to order some smaller copper shims instead.

 

Folks, I have a new Toshiba XG5 drive it seems and this is a lot cooler when gaming. 50mins BF1 at 4GHz and GPU OC 190/250 and max temp of the drive was 62C!

 

I haven't had any luck getting my core differential down on my 15r3w/1070. I changed the pads on mine like the 17w/1070 picture. My best results are with 0.1mm tape in place of the 1mm pads on the cpu side, but I still have a 15-20 degree diffference. Am I doing something wrong? I ended up chickening out on liquid metal because it seemed like I had to put too much to get a good contact. My next thought is to try the 0.1mm copper shim, but it appears that method has conflicting recommendations. Any thoughts on this? I'd really like to get my core differential down to at least 5-7.

I'd try to get another replacement heat sink (already had one), but dell has told me that it has to overheat for a swap and a difference isn't enough. Not to mention I've also had to get a keyboard replacement and wifi card replacement, so they're probably going to blacklist me haha.

 

You guys think this will work for the PCH on the AW13?

http://www.ebay.com/itm/3Pcs-Self-a...079500&hash=item2a8634be7d:g:Oa8AAOSw1~JZUoYV

 

Great to hear that you've fixed the WiFi pin issue.

As for the higher PCH temps, it is often due to the pressure from the black frame piece causing the aluminum heat spreader to sit crooked.

It's important that the heat spreader/sink is sitting flush over the PCH (die) surface.

The PCH area does get enough air flow as I've tested this before publishing the PCH mod.

Have the fins sit horizontal as shown in the pictures as that is the direction of the air flow.

I would not recommend using the self adhesive tape as those have very poor thermal conductivity properties.

It does more harm than good.

No.

The heat sinks are usually aluminum and LM does not pair well with aluminum material.

It is not worth it to LM the PCH surface, even with a copper heat spreader/sink...which I wouldn't use copper anyway do to its thermal properties.

Aluminum dissipates much heat much faster.

Even with the AW's sitting flat on the desk, it restricts a lot of air flow.

Also the surface under the fan intakes gets warm (the desk) so you're just sucking up more hot air that doesn't do the cooling much justice.

That is very good for being flat on the desk.

Air flow is important.

Take your time, have all the proper tools in advance and take your time.

I don't recommend aida64 for real world testing as it is extremely over the top and is as unrealistic as the opposite end of a synthetic bench.

The total size of the width is flexible, but as for the height around ~3-4mm.

Yes unlocking the TDP is safe as the CPU will never reach those figures.

The thickness of the pads will vary.

It is recommend to order more of the 1mm or 1.5mm to have on hand than having smaller (0.5mm) pads.

Most areas will need a size up rather than a size down.

Not all systems/heatsinks are created equal.

There are tolerances that need to be accounted for.

I always test with the rear lifted about an inch as I find that it would be doing the system an injustice by suffocating it from proper air flow.

There are little rubber feet and gadgets you can find online that work wonders...

The ambient temp plays a pretty big factor with overall temps.

Also, keeping it on max fan is ideal if gaming for a long time...

Playing OW for at least 1 hour is ideal.

Aida64 is not a really good test and I don't know why you recommend it to everyone?

You know that Aida64 stresses the system over the top and is very unrealistic to normal usage habits / load.

In all seriousness and because this is my thread, I am asking you an honest question...

"Are you recommending that people run Aida64, knowing that it's over the top and will stress the system only for you to bash on AW for your own personal agenda?"

Most would agree that running real world tests and collecting that data is more practical than running unrealistic benches like with Aida64.

Having the TDP unlocked and having higher temps are expected. There is no surprise there.

Correct. I look at avg temps for the most part because...

Hitting peaks and spikes are largely due to the fan curves not kicking in time.

This is why I take the spikes with a grain of salt and base my data collection on the average figures...

ie...It's more practical to take average temps from multiple runs (data), rather than focusing on the spikes alone, that doesn't' really tell the whole story.

I would lift the rear about an inch. Air flow is important.

Perhaps not enough paste and uneven contact points? That is usually the case.

Make sure the surface is very clean before applying thermal paste. Even a little bit of contamination will result in big temp gains...

Repasting is more important than elevating the rear.

I agree, OW is a great testing platform for real world temps..

Skylake and Kaby Lake are indeed very different.

If one split LM, they'd know as it'll throw error codes/lights from it shorting.

I don't find Aida64 to be very practical.

It is a very unrealistic benchmark app...

I along with many here would agree that taking real world values from gaming on BF1, OW, Encoding video etc... to be far more valuable than an unrealistic over the top bench app like Aida64.

Not all data is created equal.

Most people play games and work on their laptops... not just bench.

Ah yes...thanks for adding more validity to the wifi card issue and DS4.

I too have switched all my systems wifi out for the Intel 8265 and have zero issues now....

. @DeeX was also heavy into the DS4 issue a while back...

.

 

Does anyone know the PCH die size by any chance?

 

I literally agree with everything you've said

 

~3-4mm ...you may have to shave down a 4mm as I have on some systems...

 

If yoo have hardware problems (out of the box) and don't know if you should RMA or not, I mean testing with Aida64 is one of the better options out there... + Aida64 it's self is a great software for everything else. I'm Sorry if you see it this way bro @iunlock. I mean I have put up a Hell lot of useful info the last years for those with diff problems or can't find solutions by themselves.

Spoiler: Stress Testing for Beginners: Which tool should I use?

www.overclockers.com - StressTesting for Beginners!! Which tool should I use.

Edit. Is this also a bad advice?

 

Ordered 12mmX12mmX3mm 20pcs from eBay.

Hopefully it fits in the 13. I don't know if the clearance on 13 is less.

I got my 13, my friend's 13 and my other friend's 17 to do this on.

 

To each his own... All I am saying is that to get people to run AIDA64 is not very practical and AIDA64 is NOT better than real world data like playing the games that you play the most and doing work tasks that you do the most that are intensive.

@Darkhan I'm surprised you agree with thinking that AIDA64 is good?

Again to each his own...

All I am saying is that real world is > synthetic data.

Ex. Most people just run a single pass of Fire Strike for example and record that temp., when in reality one would need to have a longer duration of gaming to get real world figures. This is why I run 3DMark on a combined test Loop, while running wPrime v1.55 in the background on a 1024M pass to get as close to real world characteristics as much as possible with using a synthetic bench.

Also, I've never questioned the other bench apps so why bother even bringing it up?

While we're at it, I quote from my own experiences after having worked on numerous systems from all different brands... Just saying...

The point being made here is that Real World Data > Aida64.

.

 

Wanted to pass along a review about the 17R4 I just got back from @iunlock (07 Sept 2017)

As an attempt not to duplicate other's positive reviews, I will simply say I had the identical positive experience. From communications, shipping, professionalism, and final product. I am extremely happy. (Actually, iunlock exceeded my expectations)

Based on my specific needs, the 17R4 was the right fit. Unfortunately as we know, the 17R4 suffers from a variety of problems and in my opinion, overheating was the most limiting. (Easily hit 99C on all cores within minutes of gaming. Thermal throttling, loud fan noise, etc... and the variety of problems associated)

Called Dell what felt like a thousand times, had a tec conduct a home visit, and even had a new replacement. Nothing fixed the problems to my satisfaction.

Despite my reservations, I boxed up my $2,500 unit and sent it across the country to iunlock.

Communications throughout the entire process was excellent. Maybe I'll be the first to say this, but it was nerve-racking sending such an expensive machine to a virtual stranger. Actually, it made me extremely uncomfortable. Despite my hesitation and nerves, iunlock kept me in the loop the entire time and coached me through the process. I saw the dozens of positive reviews about his work and knew the 17R4 had tremendous potential. So because of his street cred and desiring the full capability of the 17R4, I pressed forward.

Now, I sit here with a new 17R4 overhauled by iunlock--this machine is an absolute beast. From start to finish, I give my full support for others to trust and have your machines professionally serviced.

I appreciate the hard work and professionalism.

Five Star service!
-BadROBOT

 

I never said that, Just because I like someones thread, means that some part of their thread I did like.

He spends a lot of time here helping as much or more than anyone, I respect that as I do with your post.

 

Ah thanks for clarifying.

 

I've found the most demanding test of my 13 R3 has been XCOM 2 at 1440p since it maxes out both the CPU and GPU (not that everyone has that game, but it's example of where real-world performance can be as depending as some stress tests).

 

My investigation to my 15R3 is 2.5mm max.

 

Isn't OOCT a good test? Is Prime95 more stressful than AIDA? For GPU I use Heaven.

 

Stock voltage varies with chip sample, frequency and current. There is no such thing as stock voltage now. It's all called VID, and chips can have different presets VIDs at stock frequencies depending on quality and leakage. As frequency rises, VID rises, same for heat/current.
It also matters what is considered "stock." If stock means max non turbo ratio (x29), that could be anywhere between .8v-1v for 2900 mhz.

My 7820HK's "stock" 4.5 ghz VID is 1.13v, so take that for what it's worth, and that's considered a golden chip. Not enough to make @Papusan happy, but a little less mad
Average seems to be around 1.2v for 4.4 ghz. I saw a few bad ones need 1.28v for 4.2 ghz...

 

Is there anyone in the greensboro or NC area who could help with repaste and repad for AW 15 r3?? Just curious as i may order mine soon. Thanks

 

So I followed the guide..and thought I did everything as directed with all the recommended products. Found I had to stack some pads to get the required height but did find that I could see some gaps afternoon putting the heat sink back on so I increased the height in a few places...
So I get everything together and it powers on and after a about a min and a half you can hear the left fan start to slowly kick up it's speed, like it would under load during a game. So I quickly started the Intel extreme Tunity utility and you could see the package temp quickly climb to 90c'+ the the power limit throttle kicked in
So I immediately powered down my computer.

So a few questions
1) am I supposed to put thermal paste down before applying the fujipoly thermal pads?
Seems some vids online and seen a guy do this
2) wondering if I didn't apply enough LM to the processor and there might be a gap Any way to tell?
3) I did notice a strange smell and figured it might be the pads or the LM. So again figured it was best to immediately turn off and fix before trying again.

I know pics would help, but haven't had a chance to tear it apart again. But I did follow the guide. And believe there are no gaps that I can tell

Any help is much appreciated

Alienware 17R4 7820Hk gtx1080 with the Sunon heat sink

 

It sounds as if you either don't have enough LM or you added too many pads. If you followed the directions (make sure you referenced the pad sizes for the 7820HK w/ 1080), then you shouldn't be seeing those temps.

 

I spent dozens of hours watching vids, chatting with peeps, and reading through the online step-by-step instructions for a 17R4 repaste. I made three discoveries: 1. Use contact paper and dry fit the pieces together. (some bending and twisting may need to happen) 2. Have a variety of various thermal pad thicknesses on hand (proper fit trumps brand) 3. Most experts tried multiples times before perfecting the process.

When you add up the cost of materials, time involved to learn and more time in actual tearing the machine apart plus the x% chance something could go very wrong, I thought it best to pay slightly more and have an experienced professional tackle the problem.

Cost comparisons: (good night at a bar = cost of shipping/professional repaste with LM) Prob only 5-10% of the cost you paid for the machine in the first place. Instead of learning to repaste, you may consider working on your pickup lines

My 2 cents, or....watch a few more YouTube vids and use contact paper to ensure proper fit if you're hell bent on doing this yourself.

 

Good Luck

 

Thanks, don't really have time to ship it off to someone since I'm leaving for Afghanistan for 6 months in a week or two.
I followed the specs for the 7820hk pic iunlock posted in his guide but I could clearly see gaps so had to add more pads till everything looked tight... I'll try more liquid metal. But not too much more

So thermal paste under the thermal pads or not?

 

I followed the same guide and had no issues with the sizes stated in the pic. Perhaps you thought there was a gap when there wasn't, or you needed to tighten down the heatsink. Just for clarification, you did tighten the screws down in order?