A drop in the middle, a cross, spread it with finger, or use an old credit card to smooth it perfectly? Even simple things like applying a thermal paste under your cooler can be a bit tricky sometimes. Commonly used methods are seemingly different, and if you also consider batching, there comes a question of how to it to achieve the best results.
Introduction and test procedures
As we all know, even the best cooler is ineffective without a thermal paste. There are so many combinations of how to create a thermal bridge between a heatspreader of a chip and a cooler that we thought it definitely deserves a closer look.
We have chosen and compared four methods in our tests. One of the most recommended and used is a drop in the middle of IHS, which is spread by the pressure of a cooler. This pressure is also used with the cross method that can better reach corners in some cases. These corners usually remain dry when using combination of the drop method and less paste.
It is also very popular to use finger and a foil to spread the paste on the whole surface of IHS. However, this method is often criticized by more experienced users because the layer of the paste is not perfectly smooth and that can lead to creation of air gaps. That is why people who want to avoid this use some applicator (or simply something with a sharp edge, like an old credit card or a razor blade) to precisely smooth the layer.
When you have decided which technique to use, it is time to think about the amount of paste to be used. We worked with the range of 0.01 – 0.15 ml and we used 0.1 ml, 0.066 ml, and 0.033 ml as reference values for every method.
The paste which was used in our experiment is Noctua NT-H1. We’ve chosen it because it is a part of the accessories of popular coolers Noctua and also because it has an acceptable consistency for equal spreading, despite its higher viscosity. Some pastes tend to tear, which makes it harder to create a smooth consistent layer.
The choice of the cooler was Scythe Fuma. Because of its suitable base, it was possible to see whether the air gaps reduce effectiveness of the heat transfer.
The heating element was overclocked Intel Core i7-5930K@4.5 GHz (on Gigabyte X99 UD4) with increased Vcore to 1.275 V. We simulated the burn in IntelBurnTest (7500 MB). The waste heat was around 250 W.
After the initial burn in, the tests took 300 seconds. We used high-flow industrial fans Noctua iPPC NF-F12 2000, so five minutes should be more than enough for temperatures to stabilize. It was crucial to watch if the CPU was working with the same performance the whole time. Even the smallest offset (GLOPS) could affect the final results. We repeated every test twice (new installation included) to make sure that measured values match. The intake air temperature was 21 – 21.3 °C, ensured in our air-conditioned lab.
Before we get to the comparing, we need to add that results can differ with other pastes and configurations. The reason is differences in viscosity, pressure, and contact areas of coolers. Coolers with solid block base can spread the heat differently than coolers that put heatpipes directly on the CPU’s heatspreader.
I cannot believe after all that clear research that you recommended:
“TL; DR: All in all, the best solution for coolers with a solid block base is to use the cross method with 0.03 ml of paste. Nothing can go wrong with the drop either.” (SERIOUSLY?!)
That is absurd, GIVEN YOUR OWN research. YOUR data clearly shows that a smooth surface application, at 0.1ml, gives the best result at 79.0C. The cross & drop method at 0.033ml both result in nearly 1C higher temps. The TL;DR final conclusion should be a smooth surface spread of 0.1ml. I have seen too many unstable systems due to CPUs installed without spreading the paste.
When drawing conclusions, or the TL;DR, we take into account the amount of thermal paste used for the application. It’s true that the lowest temperature is achieved with the “Smooth surface” method using 0.1 ml, but the difference compared to “Smooth surface” with 0.033 ml is negligible. A similar result can be seen with the spreading method as well, but in that case you spend more time applying the thermal paste while gaining virtually no additional cooling performance.
We appreciate your personal experience, but one important thing needs to be pointed out here: every thermal paste behaves differently, every cooler behaves differently, and every CPU behaves differently. When all of these are combined, the number of real-world scenarios becomes quite large. The tests you’re commenting on apply only to one specific scenario. For your feedback to be fully substantiated, it would be necessary to arrive at different conclusions while using the same testing methodology that we use. It’s important to keep in mind that you’re referring to a different situation altogether—and who knows how many additional variables are involved.
— “I have seen too many unstable systems due to CPUs installed without spreading the paste.”
I’m not sure whether the instability you mention was actually caused by the thermal paste application method. I personally believe that something else was responsible for that instability. 🙂
I just take the position that PTM and carbon compound pads are the future, the longevity and ease of use outweighs the cost and I’m just sticking to that instead of goop that refuses to play ball with my autism motor issues.
That could be the case. Some coolers with pre-installed TIM might even actually use PTM pads? I’m wondering how well these pads cover microscopic gaps between the CPU IHS and the cooler’s contact surface if they were applied manually—and whether they’re expected to be pre-shaped “to size” already during manufacturing. 🙂