[Guide] Laptop thermal paste and you!

The bad news is that I'm miserable, the weather is bad, and I want to curl up in my bed, cry my heart out and read Yeats for the rest of the summer.

The good news is that I've mostly tired of doing that, need a distraction, and so decided to share my polymathic expertise with you. A lot of people buy laptops as a tool, and expect it to handle in much the same way as a toaster or a TV. You buy it, and it should work in pretty much the same way for a few years. If it breaks down or starts struggling, they write it off as bad engineering, bad assembly, or cheap components. Filthy greedy corporations ruin the planet, and now also my laptop!

Part 1. WHY

This is an entirely reasonable train of thought. If it wasn't for one thing: laptops, like any other computer device with a separate assembly for the cooling array, have one component that degrades more quickly than any other. Which is the cooling paste placed between the contact plate on the cooler, and the surface of the cpu and gpu.

Theoretically, it is possible to get past this by employing a closed construction (say, designing chips mounted with a sealed layer of a heat-conducting material that a cooling array could flow a cold gas or liquid over). From a user-perspective and production perspective, nothing would change - since you don't upgrade the chips on the laptop in any case.

And although laptop-makers increasingly order mainboard and chip configurations specifically tailored to their laptop models, I doubt they imagine the increased production cost of collapsing part of the production line would be worth it. And besides, when their customers happily buy a new laptop when it gets old and stops functioning "optimally", rather than send the old one to maintenance - then it definitely doesn't pay off to invest in expensive cooling. On top of that - do most people really notice if their laptop runs at 2.1Ghz instead of 2.4? Doubtful. The idea of doubling the cost of chip assembly ten times becomes, no matter how technically awesome the result would be, a pretty silly suggestion. (One day, though - one day..)

So we have to deal with cooling arrays mounted separately to the cpu and gpu, with the deteriorating cooling paste in between the surfaces. Here's what my cooler looked like a few months after I bought it:


I've already wiped off the cpu with a lint-free rag and some clean, quickly evaporating, non-corrosive dissolver (lighter fluid or alcohol, for example - lightly press the cloth down and soak the goop, then carefully start wiping). But you can see the contact surface on the cooling array and the hilarious amount of goop they've splashed over the chip when they mounted the thing on the factory.

What happened here is that the cooling paste has "caked" - the liquid in the compound has evaporated, and we're left with a solid, rock-hard ceramic surface. This doesn't conduct heat at all - instead it works as isolation. The only heat coming off the chip now will go through the air pockets forming between the cooler and the chip, or through direct contact with the copper (for example on the edge of the chip).

Modern cpus (thanks to heat sensors wired to the logic on the actual chip) won't critically overheat because of this, and you won't see temperatures skyrocket, or chips frying, or even any instable operation, memory errors, etc. In addition, many laptop makers consciously choose "soft" temperature limit trips for the sensors that would actually allow the laptop to run indefinitely with no cooling at all (the bi-directional temp trip on intel boards, a derived variable stored as "PROCHOT", was intended to help balancing effect load on devices that use the same cooling array across several chips on the same mainboard - but in practice seems to be used to downclock chips until the hopelessly underdimensioned cooling array is sufficient).

So you won't really notice anything on normal use, when the chip clocks down anyway (rather than run at constant effect, like desktop processors). But what you will notice is that the radiator near the exhaust will have a lower temperature than normal, while the ambient temps in the laptop will be higher, and that the cpu and gpu temps go up to maximum very quickly. In the same way, it takes a long time for the chips to cool down after heavy load runs.

Which in turn is what makes the laptop's cpu, gpu or both, run at reduced speeds (gaming laptops out of the factory often is the prime candidate for this). Either immediately, or after a short burst of processor load - you score reasonably well in benchmarks, but your framerates crash after 10mins continuous running. Or more insidiously, that the end of the benchmark actually ran into the factory temp limit, while the average score seems ok - just as during normal running - thanks to that the computer isn't running into crisis territory on non-peak loads (yet).

Part of the reason why I'm writing this guide is that I've seen a lot of gamer folks for example choosing to unlock the "soft trips" on their cpu to make the laptop run faster for slightly longer bursts - when what they clearly should be doing first is a repaste.

But all laptop-owners run into this problem long before the laptop is old enough to be replaced -- typically after about a year, at the very best in fact. Best example I ever saw was a guy who was completely happy with their laptop for a year - until I repasted it, and the keyboard suddenly became cold, the fan stopped running on battery, and so on.

Some laptop makers have chosen to minimize the problems with this by replacing the cooling goop with a gel-like tape.. thermally conductive tape (see: http://www.frozencpu.com/cat/l2/g8/c487/list/p1/Thermal_Interface-Thermal_Pads_Tape.html). Which consistently has better results over time than even the best paste-job, for example because laptops are cooled down and heated up more than a desktop - during transport before you buy it, or for example when you put your hot laptop in the backpack and run out in the cold, etc. For laptops, it's even an issue during normal load because of the power balancing. And high grade cooling compounds with high viscosity, unicorn horn grind and silver in it, etc., will deteriorate quickly then. Cooling arrays on laptops also tend to have sub-par fastening - they shift more easily (even on a perfect seating - that I guarantee you will not have from the factory either), and cause air to come in contact with the goop, which then evaporates the water, and the goop cakes quickly.

For a more in-depth explanation of how cooling compounds work, how they break, and why vegemite, vaseline and toothpaste has the same thermal conductivity as Holy Arctic Silver Unicorn Horn Grind, at least for a while, see this legendary test here: http://www.dansdata.com/goop.htm

So when choosing a thermal compound for a laptop (or perhaps if you should bother with a repaste soon), consider two things - how much you're going to run around with it outside, and how high the effect/heat on the components is, how often it's running at peak effect, or what temperatures you typically end up at.

As a short pointer - a top of the line gaming laptop with dual exhaust and 90W components that typically stay at 80-100 degrees Celsius should have a cooling compound with high viscosity and high thermal conductivity on high temps. A thin and liquid but relatively cheap aluminum or zink-based compound is a good trade-off from the top of the line products, since you know on beforehand that you need to replace the goop fairly often anyway.

While a smaller laptop that often is carried in a bag, etc., can get away with much cheaper compounds with lower viscosity and optimal thermal conductivity taking place around 60-80C. Arctic Cooling MX-2, for example, is a horrible compound for a desktop, and will likely be universally panned in all cooling guides on the internet. But it's ideal for a work-laptop, and fairly easy to work with. It's also going to last longer.

For lighter net-tops, you might consider investing in a good thermal tape as well (but it's not as easy to come by, and you might need to cut them up to get them to fit the chip tops - not always ideal for the layered/best types of pads). They are not the cheapest alternative, and should obviously never be placed under a desktop cooler or in a gaming laptop. As when they stop working, they do so extremely quickly and all at once. But they will give you very good results for a work-laptop over a very long time, and it's very easy to apply. The repaste you do then might be the only paste job you do during the life-time of the laptop.

Part 2. HOW

-What you need: small phillips screw driver w/magnet, goop of your choice, lint-free cloth, additive-less lighter fuel/technical alcohol/window cleaner ( ), fine emery paper (not critical.. sometimes comes with expensive cooling goop packages), flat piece of cardboard (or use the edge of the sandpaper), matchstick, a few pieces of tape (if you're a pedant).

So you've never opened a laptop chassis before, and worry that if you pick it apart, you're never going to be able to assemble it again. Besides, support folks insist that the warranty will be voided if you look hard at the underside of the laptop. Don't worry about that - admittedly there are laptop models out there that require a full disassembly of the mainboard if you want to get at the cooling array. And certain laptop-makers actually put a seal on top of the cooling array. In which case you should perhaps bite the bullet and send the laptop for service (for example somewhere you can talk with the service-personnel, avoid sending the laptop in the mail, and give them some pointers about what goop they should use). But for the most part, the cooling array is the last thing to be fitted in the laptop assembly. So you can get at it straight away by simply removing the back plate.

Remove the battery first of all (if it's not stuck in the laptop). Do not put the laptop in sleep-mode, turn it off.

Back plates are usually screwed down and snapped in place with some retracting, bendable plastic hooks. So when you remove the screws easily (and put them in for example a set of boxes in order, so you can put the same screws down in the same holes afterwards - at least make sure the same lengths go back in the same holes) - the difficult part is peeling the thing off. Find the loose part (usually the back near the screen), and pry gently from the middle towards the side opposite of the dvd-drive. The reason for that is that the pull at the sort of two-way hooks will be more even here, and you minimize the chance of breaking one of the retractors. Avoid inserting something along the edge and then shoving it towards the hooks - usually faster, but this is the best way to break the hooks.

If you do break one, it's not a huge crisis (you can still screw the thing stuck) - but some backplates might not fit perfectly on the chassis frame afterwards.

Inside, there are a few things to be aware of.

1. The solder points around the top of the mainboard should not be touched. If you know you're a bit careless, consider resting your hands against the hdd or across the dvd-drive, and aim in at the sides, etc. But don't put a cloth across the mainboard and think that saves the mainboard - a friend worked with laptop repair for a while, and thought that was a brilliant idea. Until he twisted off and bent several of the components by tangling the cloth around the edges and spikes. Not a good idea. Other than that, the laptop can take a lot of abuse - the typical way people wreck their laptops is by tearing a power-cord out of the socket, not by stabbing the mainboard with a screwdriver, or chipping the mainboard, or something like that. The solder and the sockets break long before the print-cards or the chips.

2. The first thing you need to do is to remove the power-connectors or power-jacks. If the battery is not removable, you should remove the battery connector first. I.. personally don't do that unless I'm going to pick off ram, or dislodge chips, reconnect something to the bus, etc. -- because the connectors are often really difficult to get off, and are easy to break. And since the power drawn from the bus is very low - you have to be lucky indeed if you manage to short-circuit something. Basically, you need to touch the battery connector itself at several points simultaneously. In which case you'll short the actual battery and damage it more than anything else.

It's possible to damage the laptop that way, though. So if you can, remove the power-connector to the battery first. The power-connector can be all kinds of things, but usually is either a sugarcube block, or a long flat rail. Sometimes there's an additional film-strip to press the contact down. Or flip retractors on the sides, but that's apparently rare (costs a cent more on the ton, most likely). What you need to take care with is that you don't pull at the socket or the wires, but pull from the inside of the contact. Be extremely careful if you need to use a screwdriver - twist slightly with the side edge on the phillips iron, for example, or use a thicker flat iron - but usually you can push the contact slightly up and then out from one of the sides. Try with a finger-nail first. Resist the temptation to use a plier.

Of course, when these contacts haven't been used since the factory, and have been stuck in the laptop for years, full of grime and the plastic is hard as rock -- sometimes these contacts will need some serious force to pull off - in that case, pry between the socket and the contact, and take care that you're not pushing the contact down towards the motherboard (some sockets are open at one side, typically the side away from the motherboard - and there's a rail on the underside that will fit a rail on the socket, keeping it in place. But it'll come off (un)easily if it's lifted up slightly before it's pulled - you can see this once it's off). Use some force, in other words, but don't continue to pull if there's no effect. There's no "pop" involved here, unless the socket dislodges from the mainboard - small movements and light force, while making sure you're not pushing hard directly at the socket. So you don't end up, however carefully, breaking the solder on the mainboard by prying the socket off the mainboard.

Next, get the power-connector for the fan on the cooling array. You might also need to displace or remove things like usb-cables or antenna cables from the wifi-modem. Note the clip on some of these contacts, if you need to pull them off. If you're as pedantic as I am, fasten the loose contacts and cables with tape out of the way.

So now we should be able to get at the cooling unit. Remove the screws on the fan first, and note if there are any unused holes for screws - use the same holes when assembling later.

Then it's the screws that fasten the cooling array across the chips - typically sets of four screws on either gpu and cpu (if both are in the system). Some laptops will have plastic hooks with retractors to steer the assembly down - these can be pulled up through the mainboard with your fingers, or a light pry with a screwdriver (I mean, with a plastic wedge, of course, because we're so careful and so on) once the screws are undone. Expensive coolers might have locking mechanisms across the top of the cooler - shouldn't be too hard to figure out.

Once these are off, pry off the cooling assembly from the broadside. Avoid pulling the fan up and potentially stabbing the mainboard with the copper or the plates for the screws, etc. Blow off dust in the fan and in the radiator - usually not a problem if some is left. And you tend to get more dust into the rotor assembly if you open it, no matter how much you clean. And please do not start to clean the radiator fins with water or any solvent, really. It'll take three minutes after you power the system on before it's worse than it was. Leave it unless you know what you're doing.

Clean the chips. On my laptop, as you can see in the picture, the chip outside the contact point is covered with a thick adhesive pad with a sheet of reflective film on top - not a terrible idea, reflects heat away from the chip once it's in the cooler array. Might need to pull these halfway off to get off the goop that has caked underneath. It's not critical that anything but the top of the chip is clean, though. And goop isn't conductive, so no worries. But there will be a small effect on the temps if the entire side is covered in caked goop.

Clean the cooling array. When the copper is clean and dry and free of goop, you can -- given that the cooler is horrible quality, and full of discoloring, like on mine -- go over the center of the contact plate on the cooling array - very lightly - with the emery paper. The point with that is to only remove some of the discoloring or uneven edges (that you don't really see with your eyes), not to shave off the copper. Most laptop coolers have horrible construction, so you won't exactly destroy the cooler, or reduce it's efficiency, even if you do get deep grooves in the copper and carelessly scrape against the grain and so on. But just brush it lightly a few times.

Note: you would obviously /not/ do this on the back of a Zalman cooler, or a cooler with actually quality copper. Also do not touch the metal on the top of the chips with sandpaper - scrape those carefully with a nail to get particles off if you can't polish any discoloring or particles off with a paper wipe, etc. (Don't get any bright ideas about creating an even "scrape" across the top, or something of that sort - cooling does not work like that - see notes further down).

When the chips and array are clean and nice, it's time to apply the paste. What we want is a thin layer that coats the entire chip. Not a gob in the middle that we press out when seating the cooling array later. Achieve that by placing a small drop on the middle of the chip (or if you're using mx-2, wipe the grit off across the top of the chip). Then spread it out with the matchstick. And finally drag the cardboard (or the edge of the sandpaper sheet - I always use that because it's thin, perfectly cut, and doesn't stick to the compound, or soak up fluid. Some people seem to be fond of using their apparently inexhaustive supply of expired credit cards.. A thick grain drawing paper also works well).

It's not critical that you do the coat in one swipe either, but don't tell people I said that. And if you scrape some of the goop down on the edge of the chip top, that's better than leaving a thick layer at the edge of the chip, or missing the edge one side of the chip - or ending up with a slightly thicker stripe on the other edge.

Some people are afraid to get too little goop on the chip for some reason. Don't be. This is very rarely a problem - thicker goop (like the mx-2) either cakes and peels off if you mess around too long (say, more than 3 minutes). While thinner goop won't cake, and is very easy to just swipe across.

Basically, aim for a thin film of goop, near transparent - not an even and smooth cake topping with the same height across the plate, etc. (And this is also really easy to do on the new-fangled intel and amd chips, since the contact point on the chip is so even and smooth - back in the long-long ago, you had to paper the top of the chips to get the best overclocks and seals).

Some heroic technicians who work in certain magazines (that I won't name) have at different points made popular the idea that you can achieve a good coat by just putting a drop in the middle of the chip, and then squeezing the goop out when you fasten the cooler. This is sort of true. But even if you used the thinnest goop on the market, the coat becomes uneven, and the goop will cake very quickly after that. I demonstrated a goop-less seating that actually worked better from the first minute once compared to a paste like that. People with much physics in their education were baffled and shocked, etc.

A thin layer also avoids seating the actual cooling array unevenly (....what you're actually doing, technically speaking, is to fill the grooves in the material in the cooling array, to increase the area of the cooling assembly that can absorb heat, at the cost of conductivity in absolute terms - the conductivity of the goop is actually lower than if the metal had contact with the copper. I.e., you're not creating a pillow the heat is supposed to stream through, or anything silly like that).

When you put down the cooling array, aim for the furthest chip from the fan (steering with the four screw assembly), and put the cooling array down in the seating again, with the fan going down last. Fit the assembly in place, and screw the first screws on the fan assembly halfway down. Adjust the cooler to the screw points around the chips, and seat the screws on the chip farthest from the fan down until you get some resistance. Don't tighten them yet. Repeat for the next chip.

Screw the screws at the fan lightly down until it won't displace sideways easily. Then tighten the screws at the furthest chip until they start to resist a little. Repeat for the next chip. Then fasten the screws at the fan/main assembly.

Now tighten the screws on the outer chip, usually the cpu, quarter of a rotation at a time for each. If you do this properly, at some point one of the screws should suddenly go a bit further with a little bit less force than the others once you get some pressure down on all sides. This is when you tighten that screw an extra quarter rotation, and follow up by drawing the other three screws another last quarter rotation. You might want to reseat the entire thing if one of the screws hit the bottom of the grooves.

Honestly, this is completely wasted on the coolers in laptops.. and any chip nowadays that isn't overclocked until it crackles.. but it ensures the best seating - and will guarantee that you're not inadvertently screwing one side all the way down, and getting a tight fit with a gap at the end, or much higher pressure on one side. And spending a few minutes extra on this might make the paste last for several months longer, etc. Repeat for the next chip.

Connect the power-connectors for the fan and the power again(in that order), put back the cables where they were. Some tape preferably not made of plastic (because the glue melts and drips off) is best for fastening the cables, if you have to do that (sports-tape cut in strips, for example). Make sure no cables stick up above the highest component, and that nothing is in contact with the cooling array, or in contact with the fan, etc.

Put down the back plate. The side with the retractors, if any, goes down first. Snap it into place - if there's too much resistance, check cables and the usb contacts, lan connector, etc. Some of the metal frames can displace or slip up for no reason, or the malplaced plastic caps could be popping up from nowhere (hello, HP) - just put them neatly in where they should fit, and be more careful on the backplate seating.

Screw down the backplate using the same length screws for the proper holes again. The first time you reseat screws from the factory, you could get some uneven resistance - the residue of whatever sealant they're coating the tips of the screws with for non-permanent seating is pushed down again. So be careful and avoid accidentally forcing the screws down at a wrong angle. Then screw all the screws almost down, and tighten them around in a circle in the end, to avoid uneven pressure on the screw sockets (that tend to be seated in hard plastic, of all things). No need to force the screws down to the bottom, the plate isn't going to fall off - we want the plate to be seated evenly, with no pull at the edges of those wonderful quality plastic seatings.

And all done. Then just hope the laptop powers on again.

And remember to be patient and let it run on low/normal temps for some.. 10 minutes the first time before running any benchmarks.. and then letting the laptop cool down to normal temps while running on the fan afterwards (more important with thicker goop and thermal tape - you don't want to ruin your first expert paste job this way).

 

Wow, thank you so much for this post, Nipsen!

It happens I've been experiencing severe issues with my hp envy CPU fan not working properly and just ordered a replacement part from China. Never tried repasting or removing the heat sink before, and was a bit worried about whether I could pull this off or not.

 

Mm. It's kind of easy to start reading a disassembly guide and get overwhelmed. But the envy.. pretty sure.. also has just six screws on the back plate (and a couple under the rubber stilts). Then just pick off the backplate, and you have the cooling array without touching anything else. ..I don't remember, but might not even need to take off the service panel first either.

 

Hey - question. Came across something on the intertron.. Collaboratory Liquid Ultra -- metal-based liquid with no particles. Didn't think that was possible. Not silicone-based.. so not a grease, basically. Making it likely have an as good heat transfer rate as the actual cooler, rather than always having a lower coefficient than the surfaces around it.

Anyone tried that? Know how it handles being heated up and cooled down often, if it corrodes quickly with an uneven seating/laptop cooler type context..?