*OFFICIAL* Alienware m15 R2 / m17 R2 Owner's Lounge

I am not referring to @Papusan 's method of sanding the heatsink down flat. He's talking about flatness for even pressure and flat pressure.
That method of sanding is "Full" sanding where you start with something like 150 grit->300->600->800-1000-2000, etc, where you are actually trying to make the heatsink flat, to fix a warped surface.

I am talking about something totally different.

I am talking about "roughening" up the heatsink with 1500 grit sandpaper to create more surface tension between the LM and the surfaces. You can do that with just like 10-15 strokes with your fingers. The roughening up of the heatsink (you can do it on the CPU die also, which I highly recommend, since then you have less chances of the LM getting off or slowly being puished to the side, exposing hot spots) allows liquid metal to "stick" to the surface much better once the tension is broken. Then you can slowly (without pressure) wipe it with the applicator repeatedly. This will "work" the gallium into the tiny little scratches you created with the 1500 grit, which will help both the gallium absorption and will help the gallium to remain in place without "separating" from the copper and pooling to each other (it likes to stick to each other if the surface grip is low or it gets thinned out).

Liquid metal loves 'roughened' surfaces (like 1500 grit buffed surfaces) and sticks much better to them long term than fully polished mirror surfaces.

One thing you will notice when doing this method--let's say you had a spare heatsink and you did the 1500 grit "buffing" just for an experiment.

If you applied the LM, rubbed it in (without pressure) for about 5 minutes+, to a nice even thin layer that did not 'separate' when rubbed anymore, then you just let it 'sit' in a dust free area for a few days, then you take a lint free cloth and some alcohol (or even hot water) and wipe off all the old LM until only the silver stain is left, you will notice that the surface now feel smoother than before and the 1500 grit "scratches" are not as obvious. Why? Because the LM "worked" its way into the scratches. That's basically what you want.

As far as 'hardened uneven layers"--that's another problem. That happens because of oxidation acceleration, and that can only be prevented by firm, even pressure, or a good airtight barrier or foam dam, something to keep out oxygen from bad fitting heatsinks (It's even nice on good fitting ones). If the gallium absorption is slow and even and not accelerated by oxygen, contact will remain good and you won't see hardened oxidized layers of tin and indium after many months. You can also do the bake method on a IHS in the oven, but for bare die+heatsinks a reapplication after a few days (leaving the silver stain) may be needed if gallium is absorbed early.

You may have a 'very very slightly' hardened layer (but still liquid) on the copper and pure liquid layer on top of it, and the contact will be nice because the LM 'worked' its way into the scratches for you.

If you did this with a fully polished surface, you would find out that the LM might wind up "collecting" where the pressure is least strong, if there was ANY imperfection in the surface at all, because there would be literally no grip when the LM gets compressed and thinned out. At that point you're dependent on how firm the contact actually is there, or your core temp deltas will skyrocket.

 

Excellent writeup, very informative

I have a question. When cleaning up a heatsink with liquid metal. When do you stop wiping? It seems impossible to fully wipe all of the black liquid metal residue from the heatsink short of sanding everything away.

 

You can stop wiping when its smooth and no hardened layers are present.
If no hardened layers are there, wipe until the surface looks and feels nice and even.

It's easier to just use sandpaper if there's any hardened layers or a super thick dark ring. 1500 grit does the job there also. Sometimes you may have to go down to the copper. That's why the 1500 grit method I use works so well even if you have to do that. I used to polish with 3000+ grit but that gave bad results with longevity.

 

Again very informative and thank you for sharing your knowledge on the subject.
To grossly summarize: 2 techniks of sanding for different purpose:
1-Roughening technik (Used on every repaste, to make LM bond better to the surfaces and many more advatages @Falkentyne)
2-Full Sanding technik (used only for proven uneven surface or hardened LM, to bring back evenness to the surface @Papusan)

I will have to repaste very soon a new heatsink so i will be using your «Roughening technik». I am confident it will help.
2 more last questions (for now ):
1- Do you recommend soaking the sanding paper in water or not?
2- Base on your explanation would foam (Mr.Fox) around the die be a better alternative to tape 33+ (iunlock) or Kapton tape. Or maybe should we just put both. Tape + foam to prevent oxydation?

 

I can't comment on the water sanding thing. But when 'roughening' the CPU die or heatsink, I use it dry. Works better and faster that way. Wet seems to be slower and polish more.
What I really like about the 1500 grit rough method is, if you spend a lot of time (like 10 minutes) wiping the LM around (without applying downwards pressure), the sanding helps accelerate the gallium absorption into the copper (or nickel). And that part will get absorbed first. If that part slightly hardens, you still have the layer -above- it that is still combined with it, blocking oxidation and keeping it from fully hardening. So the tiny scratches get filled while the LM above it remains LM'eny. It's quite nice.

Think of the youtube gallium aluminum can experiments. Notice that if you apply gallium to an aluminum can, usually nothing happens? Because the outer oxidized layer that was exposed to the air prevents actual absorption. But when you scratch it up, the gallium has a direct path to the aluminum now. That concept isn't too far off what you're doing here. Especially if you aren't doing the oven 100c pre-baking/removal of LM method (high heat accelerates gallium absorption also).

And for the other question, Use both. Tape is to stop LM from shorting stuff, not from escaping (although since tape does reduce the z-height around the package, it can help slightly there).
Foam dams or another type of seal are for completely stopping LM from escaping, and can help block air also. There are other barriers that can be used that have been discussed as well (I think some sort of squished RTV layer+Kapton combo has been done, but I don't know about that)

 

I use clear nail polish in place of tape. You need to make sure that it's non-conductive first though.

 

That works too.
The killer though is oxygen getting into the cracks especially as the LM heats up and thermal expansion/contraction happens.
This is where the difference between a sanded 1500 grit roughened surface and a mirror polished surface becomes obvious. I tested this on two heatsink samples.

I took a fully polished copper MSI heatsink (as mirror as I could get it) and applied LM to it. Left it sitting for 2 weeks.
After two weeks the entire LM was completely hardened and could not even be cleaned off without sanding.

I then did the same thing with it roughened up and worked in with more wiping with the applicator also. This time, the LM was still liquid at the top! There was some minor hardening at the lower layer, but nowhere near as severe, and the reason should be very obvious--with microscratches from the sanding, the LM gets absorbed through that layer first, leaving the top layer intact. So oxidation winds up getting blocked by the top layer since the bottom layer is absorbed faster (also from the extra wiping to get it in there sooner). You can apply these experiments logically to see what happens in real world use as well. Notice that LM left exposed to air on an inert surface (like wood) gets nothing except dust on it (making it impure) but doesn't make it harden or anything.

I'm not saying that is a scientific fact or something, but one of the benefits of making your own liquid metal is being able to experiment with it without it costing you an arm and a leg.
The one thing, regardless of what you do, is to avoid excessively high temps. Temps above 80C will accelerate gallium absorption, so something like a fully mirror surface + 95C temps+ any uneven contact areas can cause very quick temp degradation and LM getting off the surface where you want it to be, while keeping temps below 80C and sealed (and a roughened up surface is even better) can create stable long term operation.

 

So I just got bored and did another experiment.

I took the MSI heatsink, wiped off the last layer of silver with 800 grit sandpaper so that all the old LM and some of the silver was removed and the scratches were visible.
I then applied some more LM, one drop, and then spent a VERY long time wiping it. It took forever and at first it didn't want to spread. And spreading would of course leave copper traces behind.

I believe this lack of proper spreading is what causes many people's temp issues. But that's just a guess.

Anyway I was sure I would need to apply another small drop, but I didn't. I just kept spreading it slowly and slowly working it through the entire surface.
Very slowly, it started to spread all over. It must have taken about 20 minutes...but finally it was finally all covered.

I then noticed two huge things now.

1) continuing to wipe it around no longer left streaks or exposed copper anywhere (i was not applying pressure--just wiping). Instead, LM 'remnants' would continue to keep it wiped and nicely applied.

2) --THIS I THINK is the smoking gun--

The LM consistently CHANGED. And it was very noticeable!

It was no longer..."liquid metal".
It now felt 'sticky', like a wet paste....and actually felt like a very very very runny thermal paste instead of what you would expect LM to look like. The stickiness was quite obvious, almost like a poor quality thermal paste in a way that is difficult to describe. It had a lot of adhesion. There was absolutely no ability of the remaining LM to "run off" via gravity anywhere. It was sticky and liquid at the same time. But it took literally 20 minutes of very slow wiping (a VERY annoying job!) to get it to that point. I am not sure if you would feel this the same way with the Conductonaut applicators, but with the Amazon mascara applicators I linked above ( https://www.amazon.com/gp/product/B00MEC5084/ ), it felt VERY different.

What was more amazing is that there was still layers left on top, despite, when I first started, it looked like there would be nowhere near enough to spread it all ! It actually looked like there was a pleasant extra amount left somehow.

I think it's very clear that such a layer would have a very high longevity. But I am not a metallurgist.

I honestly think you guys should try this and mess around and see your results. But I think that explains the huge difference in how it looks in my picture of my application, vs your guys application. It looks a LOT more like paste than liquid in mine. And that's why.

*edit*
If you feel after the 10-15 minutes of "massaging" it into a sticky type material, it's too thin and you need a bit more, just apply another small drop (CPU) or two small drops (GPU) after the massaging, and it will look like a fresh silver layer again. Just don't apply too much. The new drop will adhere to the first layer and shouldn't clump up, and boom: nice temps. The important thing is working it into the copper or nickel plated surface like I mentioned first, then if necessary, apply a second drop and then it will spread extremely easily on top of the worked in thin layer

Of course you can bake the IHS also (LGA chips), or heatsink (ask Papusan about that, not sure how safe it is to bake a heatsink at 100C to accelerate gallium absorption).

 

Ahhh, I see why I never had any issues with LM spreading even on mirror surfaces.

I've always worked in the LM very throughly to the sticky extent you described.

 

Just noticed that there has been no mention about the Aw M15/17 R2 having nvidia optimus. Is the display connected directly to the dgpu then?