M6800 Heatsink mod idea... awaiting feedback.

After overclocking the CPU on my Dell Precision M6800 ( see here how I did it: http://forum.notebookreview.com/threads/throttlestop-overclocking-dell-precision-m6800.812606/ ), now I am considering to improve the heat transfer, as that is a major bottleneck.

I've been analyzing the shape of my CPU Heatplate and heatsink and I really don't see a good way of adding an extra heatsink to it (but I would welcome any idea).

M6800_CPU_Heatsink_NoRoom by derei posted Jan 12, 2018 at 8:30 PM

I was thinking to check if I can change the plate that gets in contact with the CPU. That one is made of copper, so if there would be a way to separate it from the heatpipes and take it off, I could replace it with a SILVER plate. it wouldn't be very cheap, but it wouldn't break the bank either (approx £15 for a piece of silver 30x30x2mm).
Silver has much better heat transfer than copper, so I assume this would greatly increase the efficiency of heat collection from the CPU (in theory with almost 10% compared to copper).

M6800_CPU_Heatsink_CopperPlate by derei posted Jan 12, 2018 at 8:30 PM

Also, I've been reading about closing the gaps above the heatpipes, to maximize contact... there was no real proof if increased efficiency (whatever i could find online had questionable results), but also please let me know what you thing about the idea.
Other people used Arctic Silver Adhesive (which is a polymer resin with silver particles mixed in composition), but that one has really low heat transfer (9W/mK). I was thinking to use low-temperature solder (i found solder that melts at 139°C, which I think would be okay, as a heatpipe should be safe to operate up to around 180°C - 200°C).

M6800_CPU_Heatsink_FillGaps by derei posted Jan 12, 2018 at 8:30 PM

Anyway, for replacing the copper plate, I would first need to be able to remove the existing one and I don't know how is it connected to the heatpipe. If it's welded, I worry it wouldn't be possible without breaking the heatpipe. Anyone here who willingly or accidentally separated a heatpipe from its plate? Anything to share about how the bond is made? I would greatly appreciate.

For now this post is for seeking possible solutions and ideas... once I will decide over a consistent method, I will update this post accordingly.

Thank you very much to all who decide to add their help on this!

EDIT1: Seems that solders have generally low thermal transfer, with few exceptions which have high melting points:
https://www.electronics-cooling.com/2006/08/thermal-conductivity-of-solders/

EDIT2: Found some interesting Solders that may have higher thermal conductivity:
https://customthermoelectric.com/solder-in100-melt-point-157-deg-c-1lb-spool.html
and
https://customthermoelectric.com/solder-in97-ag3-melt-point-143-deg-c.html

EDIT3: Soldering copper with Indium or alloys that contain Indium is not always a good idea, as it seems Indium will diffuse into copper slowly, creating a very tough and brittle intermetallic layer. When talking about a vaporation heatpipe, you don't want to give it a reason to crack.
Detailed explanation here: http://www.indium.com/blog/indium-copper-intermetallics-in-soldering.php
One way to bypass this limitation would be to nickel-plate the heatpipe.

 

It's fun stuff. I'm barely more than a noob and am enjoying the challenge of something totally out of my comfort zone. Having the right equipment is important, it doesn't need to be expensive stuff.

I'm in the process of attempting to solder together a one piece heatsink with copper tube for watercooling line from separate CPU and slave GPU heatsinks. My first effort worked great on the slave part but messed something up with the CPU part that led to worse performance (~15C) due in part to some dodgy joins and not having enough of the right clips/clamps etc.

Heat gun makes them desolder and fall apart quite easily whether 140C or 180C low temp solder. The nature of copper and heatpipes spreading heat around means they tend to all fall apart at the same time so if you want it to stay together while you only add or remove one part, you will need to clamp everything together, or you may screw up the alignment of it all without realising, and all the heatpipes and rads in the world mean nothing if the heatplate is poorly aligned to the source.

The separation is quite easy, just make sure you have it sitting on something where it won't move too much once it lets go and that the heat won't burn, aluminium foil is useful here, and have tweezers/pliers on hand as things will be hot... Try and clean up the liquid solder straight away as pools of solid stuff are problematic for re-alignment and clamping.

Heat pipes puff up when they get too hot and can be flattened again gently with a hammer, they do this well before they burst so as soon as you see this back off on the heat.

I found my IR temp gun woefully inaccurate when pointed directly at metal of any sort.

Check the thermal conductivity of your low temp solder, some of the <140C stuff (e.g. Sn45Pb35Bi20) is in theory as low as 20W/mK.

I do have a cache of spare parts though, even a spare chassis and dead CPU and GPU board that I use for testing and alignment and that I don't care about things like scratches, I highly recommend you find someone selling a dead chassis of your model or a spare heatsink especially if you don't have a fallback PC just in case you mess something up.

I vaguely remember reading somewhere that silver is hard to solder to copper iirc. You may want to check this.

 

What exactly you mean by "aligned"? Are you referring that the heatpipe to stay perfectly flush with the copper plate that goes in contact with the CPU?

I've been reading about heatpipes (even ones produced by Cooler Master) to be operational up to 200°C, but I don't want to risk it, that's why I wish to work with temperatures under 150°C.
I was thinking to get an IR temp gun, I found those for "domestic use", but as you are saying you didn't have success with it, I think I will pass and think on something else.

I'm not planning to use solder with lead in composition (also lead is known for reducing thermal conductivity), but again, is difficult to get accurate solder composition especially when getting it from amazon/ebay/chinese... There is an alloy with Sn Bi and Ag(1%) that is said to have improved flexibility and low melting point (140°C?), but I need to find a reliable source... also I don't know its thermal conductivity. If you know any good alternative, please let me know. Unfortunately, low temperature is a limiter, or I could have used silver solder (but that one ranges from 600°C-780°C and is clearly a no go).

I'm not planning to test this on my actual heatsink. It will be a side experiment on another heatsink... if it fails, at least I remain with my laptop functional ) thanks for the tip anyway.

They both oxidize extremely quickly. I also vaguely remember using borax to solder on silver plated electrical terminals, indeed I need to document myself about.

If you have any practical examples to show from your tries, I would welcome them.

 

Alignment issues from solder getting liquid, subtly shifting, then re-setting.

Lower risk of course with the dell aluminium housing with cutout and channel for copper plate and heatpipe, they would be easier than the clevo I'm dealing with now which is all copper, and all parts just sitting on top of each other held together by nothing but the solder.

Pics - see my 1070 sli build thread. The big slave one is all in bits now getting cleaned up and prepped for another attempt

Another thought... 10% extra thermal conductivity will help but usually attacking the most thermally resistive parts gives biggest gains; air followed by thermal compound. More rad fin area, higher cfm fans, liquid metal in that order.

NBR user Stresstech fabbed a silver IHS for a desktop CPU P870 and I don't remember seeing any great performance out of it, low single digits temp drop IIRC

 

Oh, I got it...this is crystallization. The metal will solidify in a crystalline form and that will affect all its properties (electrical and thermal conductivity but also mechanical strength). This can also happen if the metal cools down too fast or the temperature is not high enough. I am well aware of this

I am already using Thermal Grizzly Conductonaut and the bottleneck is the heatsink/heatpipe. Heat dissipation at the cold end is rather okay and it will be improved later if needed, but for now I noticed the heatsink can't "collect" the heat fast enough from the CPU, so I am trying to improve that. On maximum stress, I can see constant temperature growth. The blown out air goes around 50 degrees and I wouldn't mind to see it higher.
Also ,if you will look at the images in the first post, you will see the Dell Precision M6800 CPU Heatsink has two branches. The long one has a much lower thermal transfer. The hottest one is the short one, which has its own dedicated fan. The long one shares the fan with Graphics.
I was also thinking on wrapping the pipe in some insulating tape, in an attempt to increase phase convection. Doing so, I expect there to be more heat transfer at the cold end, which may increase its efficiency.

Copper and Silver have very close thermal conductivity values. Actually, in some cases, copper can be better than a low quality silver. Problem is that I don't know how good the copper in this heatplate is, as overall is full of imperfections (dell manufactures it to be cheap). So, that may be one reason why there wasn't a visible temperature drop.

 

In your case it's less about the heat transfer and more about the mass. You need to improve on the actual heatsink not the heatpipe. Your heatsinks cannot collect enough heat and your fans can not cool down either. So before making thoughts about how to optimize the heat transfer you should first improve on the vaporation part. There is no point conducting heat if you can't get rid of it.

Also you can remove the heatpipes with 140c, important to note is that you do not heat it beyond that point, other than you want to burst your heatpipes open.

You have to work with liquid solder and make sure the melting point is 138c, there are liquid ones around 200c which would not work with your heatpipes, in case you want to later on when making bigger heatsinks,add some thicker heatpipes.

 

@Danishblunt what would you do in my case? Rebuild the entire system with bigger heatpipes?

 

Don't bother with a silver IHS. Assuming you build it right you'll get a 1C improvement at best.

Your weakest link right now is the radiator size. The radiators are too shallow for the air blowing through to pick up much heat. Try to fit on thicker radiators.

If you really want to add an extra heatpipe, just stack it on top of the two existing pipes.

 

I don't really see how that could be achieved, as the radiator is fitted tightly between the laptop chassis and the fan. It could be a way to trim the fan's case, so a very slim radiator would fit in between (a radiator also cut from a second heatsink assembly and then trimmed into half, somehow), but that would require to collect heat from the heatpipe... which would mean a secondary heatpipe to run from it (but this is already stated in first post: I don't really see a way to fit that extra heatpipe, as there isn't much place - on top can't fit, as the palmrest won't close anymore).

See here how this laptop looks disassembled and how heatsink assembly fits in: http://www.laptopultra.com/guide/dell-precision-m6800-disassembly-pictures/
Also, here the User Manual from Dell: http://www.dell.com/support/manuals...14a311-8a1f-4fb1-b59d-701bd5e0d605&lang=en-us

As you can see, there isn't much room to play with... is not like other laptops, with the heatsink running loose by the middle of the mainboard. The pipes and the radiators are quite snug fit and there isn't much room to play, which limits my options considerably... Also, if you will pay good attention to the plate that presses the heatspreader, you will see the heatpipes have their own grooves in which they sit, which makes quite difficult replacing the heatpipes with bigger one, as they won't fit in their place anymore.

 

I made this video a while back. It shows use of the 138°C Chipquik solder paste to solder a copper shim to a heat sink. I have done a few. @Mobius_1 is using one of them on his P870KM1. When assembling it is better to use an oven than a heat gun. Temperatures are better controlled and more even. As @bennyg mentioned, you cannot have too many clamps. Clamp everything and clamp the clamps, heat it up until the temps are achieved and the solder flows, then shut the oven off and let it cool off before disturbing it. @Khenglish is the master of masters with this kind of stuff.

With severe chassis constraints to contend with on laptops that can barely even accommodate what little they already have, it's going to be very difficult to achieve meaningful thermal improvements. Everything is pretty much engineered to be marginally functional and squeak by as something that is almost good enough by the skin of it's teeth. Most everything that is designed to appeal to the thin and light crowd does not even have a smidgen of wiggle room for improvements. If it did, then it would not be appealing to that group because it would be too thick, heavy and robust.

At least your M6800 has MXM, so that's awesome. http://www.dell.com/support/manuals...d61488-0c5e-459e-80a0-eba03a606abe&lang=en-us. These are really gorgeous machines that look just like the XPS line. My wife has an XPS 15 and loves it. It looks awesome and it is made well, with tons of aluminum and carbon fiber lower cover. But, it leaves a lot to be desired in terms of being mod-friendly. There is no extra space for anything. It is only engineered for people that merely use their notebook as it ships from the factory. It has no capacity for fancy tricks or acrobatic feats.

 

Don't put two radiators in series. The misalignment in the fins will cripple air flow. Try to find a little bit thicker one and replace the existing radiator.

You can also make radiators denser. For how thin yours is you can disassemble your radiator and sand the fin edges down, then reassemble at around 50% higher density. You need to have a 2nd radiator to also break down to maintain the length. When I did this I got a 6C improvement, and I think you'd get even more. You will need to dust your laptop more often, which is why I think the rads aren't made thicker to begin with.

 

As stated heatpipes are not really your issue right now, it's the heatsink size. You need more fins and mass and possibly a better fan.

Sidenote on what kenglish said:
You can add more fins, but don't add to many since you will worsen your airflow and make it worse. So make sure you still get a decent airflow.

 

Yes, that is the reason why I started to document myself about improving it. I replaced my video card from FirePro to 980M which made a huge difference. That was the first time when i opened this machine.
Funny thing, this machine is heavy like hell (4kgs) and thick (~3.5cm), still the cooling is designed to allow too little.

Thank you for the video! Extremely instructive. And for the oven tip.

 

On the first post ( http://forum.notebookreview.com/threads/m6800-heatsink-mod-idea-awaiting-feedback.812796/), second edit I added some Solders that I just found, which seem to be better in terms of thermal conductivity (at least Indium is advertised to have 80W/mK according to Wikipedia: https://en.wikipedia.org/wiki/Indium)

 

Liquid ones are much better so you can spread everything properly.

https://www.ebay.com/itm/100g-Solde...e=STRK:MEBIDX:IT&_trksid=p2057872.m2749.l2649

I used this for custom heatsinks.

 

I totally agree with you especially when soldering flat surfaces. I wrote to them asking if they provide those as paste... we'll see.
Do you know anything about thermal conductivity of this paste? Whilst it should have a very thin layer only (hopefully, if anything goes on right), still a low thermal conductivity so close to the CPU could have negative effects.

Edit: nvm... found the thermal conductivity of that one: 19W/mK ( http://alasir.com/reference/solder_alloys/)
screenshot here: https://imgur.com/ztppE9i

 

It's solder, it's high enough for you to be satisfied. As I said, your focus really shouldn't be on conductivity right now. As me and kenglish already told you it's way more about your heatsink and fans. You need to make it bigger, add more fins, add a different fan into your case etc. Even if you were to replace your heatpipes with 3x thicker ones, you'd have next to no impact on temps, since the bottleneck is entirely on your fan / heatsink.

 

Hi
I do agree that solder is't Your problem, it's so thin layer of it that it not affect conductivity much. Besides it's used for joining heat sinks with heat pipes and radiators in factories for a reason, because it's the best solution. Much, much more gains can be achieved by focusing on radiators and fans. It simple, more air fan moves it's better. Finding fan with higher CFM and replacing old one is good direction.About radiator, I don't know if You look actually on it but it is different from type which, I think Khenglish, was modding. The heat pipe is soldered directly on top of it. That's why making it denser isn't option. There's really no way to keep those separate fins together during soldering, or at least it would be extremely difficult.
That's, what I think, looks like cross section of the radiator: But derei have heat sink "in his hands" and can confirm or show us how it really looks like.


Replacing radiator part with copper one would help for sure as well but it's matter of finding one. There are some DIY copper heatsinks on the Aliexpress but they would need to be modded off course to fit them into the chassis on the place of aluminium one.
But I don't agree about conductivity. I don't know on base of what You draw Your conclusion that more heatpipes or bigger one would not improve temps. If I look on the heatsink itself, immediately what stands out is that for a look of it , closer radiator on the right is connected to the cpu plate with narrower heatpipe. It doesn't make sens for me. Why, judging by the eye, radiator with two times more surface area is connected with smaller heatpipe? I would look into this area. It would took quick test to see if extra heatpipe would help there. Take out motherboard from chassis and check under load what are the cpu temps. After that source some short flat heatpipe, either from desoldering old laptop heatsink or buy new one and put on the top of heatsink or even stack it on the existing heatpipe connecting cpu plate with actual radiator. Use maybe very thin thermal tape or thermal paste to allow better heat transfer and repeat stress test. That could give You some insides if the limits of current configuration are actually there. It's not the best and accurate way but quite easy to do. Anyway that were my two cents, maybe will be helpful.

 

@darnok44 thanks for your input. Soon the new heatsink assembly will arrive (I ordered a spare one, as I don't want to risk damaging my existing one and have my laptop non functional). Once I have it in hands I can work on it and see what can be done.
First steps will be to do some thermal transfer tests, so I can have some results. Then I will also look into your idea of getting more contact surface between radiator fins and heatpipe. Indeed, at the moment it's only connected at the bottom, so the contact area is rather small. Your suggestion is interesting, but it depends on how much I can modify. And of course, I want the modifications to be within a limit, so it would be practical to others to follow it if they wish so. If I redesign the heatsink (almost) entirely, then it may be too discouraging for others.

 

Actually my main idea was to check if this smaller shorter heatpipe isn't a bottleneck in this heatsink. And if this would be the case, add second heatpipe above existing one, which not only possibly help cool cpu die but as swell VRM's and chokes which are above cpu and covered with the heatsink to. I didn't think about that in that way that it will increase contact area between heatpipes and radiator but when You bring that up...yes, adding second heatpipe would do that also.