I’m still on the fence about whether or not I should keep my MSI GS63VR or sell it and get a smaller, lighter laptop for travelling. It’d be sad to part with it, as I’ve invested a fair amount into upgrading the laptop (new M.2 drive, new HDD, new display) and people generally don’t like buying laptops that the user’s opened up. (Based on my initial attempt to sell the laptop, anyways.)
Having watched quite a few tutorials on applying liquid metal (Gamers Nexus delidding videos, etc) I’m pretty confident I could apply the liquid metal without any issues. And after watching LTT’s recent video on applying LM to a laptop, it’s hard to believe I could screw things up more than they did on their first try! But those thermal results were pretty outrageous.
I’ve already opened up my laptop and re-applied the thermal paste once, I used Thermal Grizzly Kryonaut, and saw 5C to 10C lower temps–depending on the core–using Dolphin’s CPU benchmark. (I used that as my benchmark and test my undervolt because that was the kind of workload I was aiming for.) So, as far as disassembly and application is concerned, that should be a walk in the park.
The issue is, I don’t know how to be sure what material my heatsink is made of and select a suitable LM compound. Furthermore, I don’t know whether a clear barrier (like the nail polish GN prefers) around the CPU/GPU dies will be adequate to protect my entire laptop. Can anyone help me understand the risks of LM scattering inside my machine and help provide suggestions for managing them? I’ve heard you can use electrical tape to help create a barrier, could I just make a rectangular “trap” around my CPU and GPU and be safe?
Thermal Grizzly Conductonaut seems to be the best LM compound out there at the moment. Haven’t seen one post about it degrading over time.
From looking around online, the heatpipes and CPU/GPU blocks of the heatsink are copper, and the mounting hardware and memory/VRM block are aluminium. There was a post on Notebookreview’s forums I read a year or so ago, in which the user had used a high temperature epoxy (think it was JB Weld) to insulate the aluminium parts of the heatsink on their Razer Blade. A lot of people have had luck with using 3M Super 33+ tape to insulate the SMDs around the dies - as opposed to using nail polish like GN do.
The risks of the LM spreading into other parts of the machine is quite low, as long as you don’t apply too much, as it has a lot of surface tension. Tape isn’t the best thing to use to completely close the gaps around the dies, as it can’t compress. You could use some kind of tack material, or high temperature, low density foam to create a barrier around the dies, if you were that concerned. If you could use tack to gauge the distance between the chip and heatsink when it’s screwed down, you could also use thermal pads to create the wall.
Not to make a confrontation, but are you sure you did things correctly?
Lots of laptop resellers offer liquid metal application as one of their customizations, and they still offer warranties on the units they apply the LM to.
Well a lot of laptop resellers are dumb, just the world we live in. It takes a while for liquid metal to leak. Even if you apply too much, don’t nail polish the capacitors on the package, don’t tape shut the heatsink-cpu, and carry your laptop around with you, it could take months before the liquid metal shorts anything to the point that it becomes apparent.
A good application of LM may take a year before you encounter problems, by which time you will be out of warranty anyway.
Thermals are great though, reduction around 8-9C over high-end thermal paste. The difference is greater than TP vs LM on desktops with good heatsinks. This is probably because of higher mounting pressure on desktops.
I would avoid using LM completely for every single use case, it’s a gimmick that is going to kill your equipment. Yes you may see slightly better thermal transfer then a good paste, but at what cost?
LM is really Galinstan, not only is a risk due to it’s conductive nature, is has “aggressive corrosive properties (it corrodes many other metals such as aluminium by dissolving them)”. It doesn’t matter how well you seal things with nailpolish, it will eventually in all cases eat it’s way through and damage the CPU, there is no doubt at all that this will happen in all cases eventually.
While this video is about mercury, it’s the exact same process and has the exact same effect, no amount of “cleaning” or “sanding” will remove the Galinstan once it has been applied, it will continue to amalgamate until it’s completely compromised the metal.
If you really want to get optimal transfer and are willing to void your warranty, or potentially damange your CPU, there are three things you can do.
Lap the CPU heatspreader until it’s flat and has a mirror finish using a flat surface, a cloth and some fine cutting compound. A bit of elbow grease goes a long way.
Lap the heatsink to a mirror finish using the same method.
Use a decent thermal paste, not LM.
I have done this many times before and in testing it is on par with using LM.
The thermal conductivity of liquid metal increases by a considerable anmount as your components get warmer, so if you’re going for maximum performance(and perhaps longevity) it’s not that bad of an idea.
you are better of just applying conformal coating, as in most cases the electrical tape or whatever you might want to glue to your smbs pushes your cooler off the heat source.(not to mention that smbs are really fragile)
This is completely incorrect, if anything LM causes issues with thermal conductivity as it amalgamates and compromises the crystal structure of your heat spreader and copper heat sink. The amalgamation of copper and Galinstan has a far lower thermal conductivity then that of pure copper.
It certainly is, if you’re willing to risk your CPU why not actually setup a proper custom water loop with enough capacity to actually carry the heat away. I am running a Threadripper 1950x @ 4GHz, no issues at all, never exceeding 30C using thermal paste, a lapped surface and a water loop. The loop also runs through a 1080Ti, even at full load the system never exceeds 30C simply due to the pump, radiator and thermal capacity. Again, LM is a complete gimmick, the tiny benefit it gives is simply not needed nor worh the problems it causes.
Untrue, you don’t seem to get that it amalgamates, which means it enters the metal structure and creeps through it like an infection, if there is enough there (which there certainly is) it will eventually creep under the spreader and make contact with things it shouldn’t. It amalgamates with Gold too, which your CPU uses to connect out to the grid array underneath (bond wires).
This thread is about laptops. I don’t know of many that have CPUs with an ihs. The contact plate for laptop cooling systems can be AL but I mostly see copper pads for the die contact area which is far less reactive to LM. I’m not sure why you keep mentioning an ihs and also mentioned custom loop and TR. Apples and oranges.
The galinstan shouldn’t react with relatively inert organic materials, like epoxy resins. Would be a much safer choice, but you’re also losing any chance of a warranty being honoured, because you’re not getting the epoxy off without damaging the chip.
I’ve seen people reporting liquid metal pastes staining/pitting the CPU dies, but beyond that, there’s been no clear electrical damage. The silicon shouldn’t react chemically at normal operating temperatures. Conductonaut appears to not be as offensive as other liquid metal pastes, when it comes to this.
This is why liquid metal use is so popular in laptops with enthusiasts. The thermal capacity of the heatsinks is too low in the first place, so the only real way to improve thermals is to improve conductivity in all parts of the system. Uneven and weak mounting pressure, and the lack of an IHS all contribute to bad temperatures. iUnlock has done plenty of mods where they’ve lapped the heatsink to improve temps, and they’ve shown that even after that, using liquid metal shows an improvement, however minor. Their Aero 15 mod is a good case to look at.
The rate of amalgamation has been shown to be relatively slow with copper. People have been using inferior LM pastes for years with no major issues. By the time that it’s had a lasting effect, chances are you aren’t going to be using that machine anymore.
More of a risk of the liquid metal killing a CPU by shorting it, before the pins are amalgamated.
Obviously, it’s not for everyone. Repasting would need to be done more frequently, and with greater care, but I don’t think it should be dismissed completely, if you know the risks.
While IHS and custom loop don’t really apply to a laptop which is obvious, I was trying to show that the paste is not the weak point in the cooling setup, it’s the ability to carry the heat away from the CPU that matters most. Too much stock is put in heat sink compounds, it obviously plays a role but resorting to exotic materials rather then just lapping/fixing the interface is a bit of a last resort hack.
This is because it amalgamates with the silicon, which is a metal. Leave it long enough and it will cause issues, although it may be many years before it has a detremental effect.
If you’re not already aware of it, the iPhone helium exposure issue is caused by a lack of a protective metalized layer on the MEMS oscillator die, allowing the helium to diffuse into the silicon. It is highly likely that the pitting is compromising the protective layer that is used on the die for a CPU, which indirectly could cause premature failure of a CPU.
Agreed, which is why I am quite perplexed that people put it in a laptop that will see a ton of movement, vibration, flex, etc… It has a very high chance of getting somewhere it shouldn’t.
This is the point I am trying to make, yes it has an improvement, but compared with the risk associated and the cost of a little elbow grease to lap a heat sink, why take the risk?
It would be difficult to lap the contact plate well. For most it would require things they don’t have access to or skills they don’t have. LM offers a competitive alternative.
The hardest part would be lapping the interface on a strangely shaped heatsink/pipe. Yes this would be time consuming but as far as equipment and skills go, it’s not hard at all, or expensive in any way. I stand by my statement that LM should be considered as a last resort where lapping is impossible for whatever reason.
It’s not exactly hard to rub in a figure 8 pattern on a flat surface with wet & dry, and then a cloth and some cutting compound, it just takes time.
Edit: Perhaps I am just frustrated by how often I see people bring me broken PCs because someone told them to apply LM to it.
Would think to would be near impossible to get a flat surface from the contact plate because of it’s shape as a whole. It would have to be desoldered from the heat pipes and then lapped IMO.
Silicon isn’t a metal. Silicide formation is highly endothermic, so I doubt there’s a chemical reaction happening between the liquid metal and the silicon. Maybe the pitting is caused by larger-scale processes? Something like copper in the liquid metal not being completely amalgamated, and physically wearing the surface of the die via friction? Pretty neat is someone did some research on that.
Helium atoms are tiny in comparison to those in liquid metal, and they don’t have to deal with mechanical forces like surface tension. The helium’s inertness is the problem, which is completely the opposite problem here.
Your goal isn’t to get it level, but flat. The mountings especially in a laptop will have skew/twist tolerances to allow for a non perfect heat sink, which is partly why they are mounted with springs. The goal is to remove as much pitting as possible so there is more direct metal to metal contact.
Incorrect, sort of… it’s a metalloid as it doesn’t quite fit the definition of a metal or a non-metal.
Perhaps so, but whatever the cause the pitting as described would damage the protective layer the silicon is doped with.
Correct, I was not inferring that Helium would cause an issue here, it was simply to show that this coating is to protect the silicon and a lack of this protective coating can have unforeseen consequences.
Every time someone thinks of using LM or suggest someone else to use it, it just tells me that those people have no idea what they are about to do and what the dangers are.
I’ve run tests with ridiculous thermal compounds and can agree with this chart as I too found that mayonaise worked kinda well and that chocolate was a terrible for thermal transfer, so if you really want the best results you need to lap the heatsink and apply some kind of thermal compound.