Hacker News Reader

> since modern thermal pastes are non-conductive, there is no real downside to an enthusiastic application of paste except for making a mess, so you want to default to putting too much on rather than having some dumb intern kill a PC with an insufficient application. There is virtually no effect on thermals.

The effect on thermals are absolutely not insignificant. Thermal pastes are actually extremely bad at transferring heat and too much of it really has negative effects.

Yes, it is safer to put too much than too little. Using too much the worst case is just that the CPU throttles and performance is sluggish and that rarely leads to an RMA.

From https://www.ekwb.com/blog/thermal-compound-guide/

For example, the thermal conductivity of a high-grade thermal paste is 8.5 W/mK, and the heat conductivity of copper is 385 W/mK, or for aluminum 205 W/mK. You can see that the thermal compound is actually a poor heat conductor and that is exactly the reason why you only need a very thin layer of paste to fill the micro imperfections between the IHS and heatsink.

> For example, the thermal conductivity of a high-grade thermal paste is 8.5 W/mK, and the heat conductivity of copper is 385 W/mK, or for aluminum 205 W/mK. You can see that the thermal compound is actually a poor heat conductor and that is exactly the reason why you only need a very thin layer of paste to fill the micro imperfections between the IHS and heatsink.

This has always been my rule of thumb whether it's metallic or non-metallic paste. I throw on a pair of vinyl gloves, squirt a dab on the heatsink (not the CPU), and spread an even layer with my finger. I go back and fill in low spots with a smaller dab, trying for an even surface. This way, it's thin enough that it doesn't squirt out the sides when it's installed, but it's a uniform spread that covers any gaps between the heatsink and the CPU surface. Also, by putting it on the heatsink instead of the CPU, I don't risk accidentally getting paste on the non-contact parts of the CPU.

I've also used the semi-solid thermal pads similar to what comes from the factory on most OEM heatsinks, but they are far less efficient than paste.

For what it's worth, Arctic Silver recommend not to use your finger, on account of dead skin cells and grease.

Page 4, http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_met...

To be fair, the Poster did mention they wore gloves.

Right, hence the vinyl gloves. It also protects my skin from any harmful chemicals or minerals that may be in the paste.

so IIUC thermal paste is only 'better' than air gaps :)

anyone ever friction soldered a heatsink to an IHS ?

This is possible but friction welding needs temperatures to reach above melting point for a little while that is 1085°C for copper and 660°C for aluminium, machinery needed to do this would be insanely expensive and would need to be done to the IHS detached from the cpu as that gets ruined at around 115°C i think, it would be better for enthusiast grade heat sinks to carve out an IHS in their blocks. but that would be made for mainstream processors only 8700k, some intel hedt and amd threadrippers

Edit: selling custom IHS integrated heatsinks with LM applied and nail polished is a small and probably fairly inconsistent business to invest in buying a well binned 3790k kit, very few people would upgrade till an 8700k which showed significant gains worthy of an upgrade you would run dry for a few years especially the premium you would ask for such a service people would use these things for a long long long time i know people who are still using i5 2600k just because majority games are poorly multithread optimized

AMD could collaborate with EKWB would be great TR2 Liquid and Navi 96 Liquid. Intel have too much ego to collab

High mounting pressure will squeeze out excess thermal paste which greatly narrows the difference between the most and least conductive compounds when both surfaces are fairly smooth and even in while in contact.

https://www.overclockers.com/heatsink-mounting-pressure-vs-p...

> since modern thermal pastes are non-conductive, there is no real downside to an enthusiastic application of paste except for making a mess

I recently re-applied thermal paste (an old Arctic MX-2 I had lying around) on a late 2013 MacBook Pro which has bare-die CPU and GPU. The first time I did it, I applied too much paste and it had an observable negative effect on thermal dissipation. Once I applied the correct amount, I got slightly better results than stock paste.

That's strongly mitigated by mounting pressure which is very high in macbooks.

> there is a very real danger to insufficient paste application on a bare die

So, I've repaired and cleaned many, many laptops and computers over the years. For bare die components (CPU/chipset) I use something like "half a grain of rice per finger nail area". This is already plenty. I sometimes checked after mounting, and there was always some minor squeeze-out. (For desktop CPUs I also use just half a grain of rice. The higher mounting force results in about a ~3-4 cm dia circle; this doesn't extend all the way to the corners of the IHS, but it doesn't have to, since nothing is dissipating heat there.)

> there is no real downside to an enthusiastic application of paste except for making a mess

That's not my experience.

With laptops, the pressure applied by the cold plate is not that much and thermal pastes have high viscosity. (The cold plate is often not that stiff, too, so it tends to flex). Applying too much paste results in a too thick layer of paste, which results in higher temps, hotter laptops, fans spinning faster and thus more noise and annoyance.

Indeed often I found the issue with manufacturer-applied paste was that it was simply too much. I think the reason they do this is that it's harder to get wrong and it makes the product just more annoying, not break it. If they did it properly, there would be some chance of getting it wrong, resulting in a broken product. For the same reason they're using thermal pads wherever they can get away with it: They're practically impossible to misapply.

> if the thermal sensors are covered and cannot detect this

Modern chips seem to have sensors specifically detecting hotspots. E.g. with an AMD Polaris GPU, you can see "Hotspot Temperature" in HWiNFO.

Yes but they might not have as high a density of sensors at the periphery/uncore.

I think that might be the reason indeed. It's better to err on the safe side.