4 ms·

If a bare-die CPU or GPU is not fully covered, the uncovered portions will rapidly become hot and the chip may die (especially if the thermal sensors are covere

by paulmd 8y ago

If a bare-die CPU or GPU is not fully covered, the uncovered portions will rapidly become hot and the chip may die (especially if the thermal sensors are covered and cannot detect this). This is different from modern desktop processors where there is an IHS to help spread the heat around - there is a very real danger to insufficient paste application on a bare die. In contrast, 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. https://www.youtube.com/watch?v=rAid5G30-WM https://www.youtube.com/watch?v=rAid5G30-WM LM is conductive, so you need to be much more careful with the application, as over-application will kill the chip. It also can migrate over time, especially if you put a bit too much on, and is particularly unsuitable for laptops that are going to be moved around a lot. Putting LM on a laptop is a gimmick, and one with fairly considerable risk to the hardware, not something that is suitable for mass production.

aequitas 8y ago
I think that might be the reason indeed. It's better to err on the safe side.
lvillani 8y ago
> 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.
- andromeduck 8y ago
  That's strongly mitigated by mounting pressure which is very high in macbooks.
tjoff 8y ago
> 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/ 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.
- morganvachon 8y ago
  > 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.
  - MaxBarraclough 8y ago
    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_method_vertical_line_v1.1.pdf http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_met...
floatboth 8y ago
> 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.
- andromeduck 8y ago
  Yes but they might not have as high a density of sensors at the periphery/uncore.
blattimwind 8y ago
> 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.

by paulmd 8y ago

aequitas 8y ago

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

lvillani 8y ago

> 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.

andromeduck 8y ago
That's strongly mitigated by mounting pressure which is very high in macbooks.

tjoff 8y ago

> 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/ 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.

morganvachon 8y ago
> 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.
- MaxBarraclough 8y ago
  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_method_vertical_line_v1.1.pdf http://www.arcticsilver.com/pdf/appmeth/int/vl/intel_app_met...

floatboth 8y ago

> 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.

andromeduck 8y ago
Yes but they might not have as high a density of sensors at the periphery/uncore.

blattimwind 8y ago

> 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.