Noctua NT-H2 Thermal Compound Investigation: Re-Test of Pastes from 2023 and 2024 – Series Dispersion or one-time Slip-up?

1 - Inbtroduction and overview 2 - Thermal resistance and bondline thickness 3 - Thermal conductivity, GPU and CPU performance 4 - Degradation, maturing and outgassing 5 - Micoscopy, particles and material analysis 6 - Summary and conclusion

Degradation in the first cycles and possible trend

Thermal paste can vary in its thermal conductivity during the first few cycles of heating and cooling and can even lose this in some cases. This happens mainly because the paste is mechanically compressed when it is squeezed between the uneven surfaces of the processor and heat sink. This process can initially even improve conductivity for a short time, as air bubbles are reduced, but sometimes also leads to small gaps, which then permanently impair thermal conductivity again.

In addition, some thermal pastes contain volatile compounds that outgas during initial heating. This outgassing can lead to a reduction in material density and thus to a further decrease in thermal conductivity. Repeated heating and cooling can also change the consistency of the paste. It could harden or settle, which impairs the ability to conduct heat efficiently. In addition, thermal cycling will cause the paste to age, leading to a long-term deterioration in performance, particularly with inferior products. However, high quality thermal pastes are designed to remain stable over many cycles, so this effect is minimal.

I invested a whole day in this analysis, which accurately depicts the initial behavior but can only show a trend for further aging, because I think you should definitely do this at least once. The industry tests 1000 cycles and more, but at shorter intervals. Nevertheless, a trend reversal is no longer possible after just 5 intervals, so I can definitely make a reliable statement here regarding the initial behavior of a paste. The Noctua NT-H2 loses around 4 percent of its effective thermal conductivity at 100 °C in the first 7 cycles, which is rather mediocre. All DOWSIL pastes from the tests are only around 1 percent here, but there are also significantly worse pastes such as the Thermal Hero Quantum. When cold at 25 °C, however, the Noctua NT-H2’s reduction is already over 40 percent!

Since the normal temperatures on a CPU die are well below 100 °C, the first 2-3 values are actually useless here and I would advise every colleague to only test this paste after at least two complete and longer load and cooling cycles. I know it takes time, but I also do this before I test pastes. However, each time I do an outgassing cycle at around 150 °C and 500 µm BLT so that the paste finalizes like a good dough.

The matrix used and the outgassing

The outgassing of siloxanes with increasing pressure is a complex process that is influenced by the interaction of pressure, temperature and the physico-chemical properties of the siloxanes. An important factor is the increased solubility of gases in the material. With increasing pressure, gases dissolve better in the material structure, and with a subsequent increase in temperature, these gases can escape more quickly. In addition, the higher pressure leads to compression of the spaces in the polymer matrix in which volatile components are stored. When heated, these molecules become more mobile and can be released more easily. The pressure also changes the physical properties of the material, making it less flexible and less able to retain volatile molecules.

Increasing pressure can also influence reaction kinetics and accelerate decomposition processes, resulting in more gaseous products. Another factor is the increased vapor pressure of certain volatile components at higher pressure, which in combination with an increase in temperature leads to increased outgassing. These mechanisms explain why the outgassing of siloxanes can increase with increasing pressure, especially in combination with an increase in temperature in the range of 75°C to 85°C, as we can see here in the next graph based on the abnormally increasing BLT at 40 N, which normalizes again to the value of 4 N above 90 °C.

But how does the “hump” come about, which is not visible at 4 N (corresponds to the usual maximum 40 N for larger graphics chips), but which is extremely pronounced at 40 N (more in the CPU area, Intel LGA 1700)? Unfortunately, I can’t analyze organic compounds in detail, that would take me too far afield. However, the viscoelastic behavior does allow certain conclusions to be drawn about the siloxanes used. A food chemist friend of mine made an interesting comment on this some time ago.

The outgassing in a possible compound with acetic acid (acetates) could be explained by a similar behavior, especially in the temperature range from 75°C to 85°C and with increasing pressure. Acetates, which are used as crosslinkers or catalysts in certain siloxane formulations, may contain volatile components that lead to gas release under certain conditions. As the pressure increases, these volatile components, such as acetic acid compounds, dissolve better in the polymer matrix, and as the temperature increases, their mobility increases, making it easier for them to outgas.

The pressure compresses the spaces in the material in which the acetates are embedded, which causes these volatile molecules to escape more quickly when heated. At the same time, the pressure influences the physical properties of the siloxanes so that the material is less able to retain possible acetates and other volatile components. In addition, increasing pressure can accelerate the reaction kinetics and decomposition processes of the acetic acid compounds in the siloxanes, which leads to an increased release of gases. The increased vapor pressure of the acetates also contributes to the fact that outgassing is particularly pronounced in a certain temperature window when pressure and temperature are combined.

Generally speaking, if you look at the behavior when heated, an increase in temperature leads to thermal expansion, as is the case with most materials. The layer thickness usually increases. If the material is not yet fully cross-linked, heating can accelerate this process, leading to densification and possibly subsequent shrinkage of the layer thickness. Once the cross-linking process is complete, the layer thickness remains relatively stable. Finally, let’s compare the NT-H2 (blue curve) with the Thermal Hero Quantum (white dashed curve), which is poor in this respect:

The NT-H2 doesn’t look quite so bad here, at least when it comes to outgassing you can make your inner peace here in terms of the BLT’s behavior. The white line of the Quantum shows how a paste virtually executes itself in the first 7 cycles. The Noctua NT-H2 remains relatively stable, despite its very liquid consistency. The behavior is certainly not ideal and you can clearly see the age of the paste. If you consider that companies like Dow Chemical alone need many years and a lot of money just to develop a better matrix, then newer materials naturally have a clear advantage. Time does not stand still. Noctua should perhaps consider increasing the degree of filling and modernizing the siloxanes used in order to remain fluffy and not end up with a rock-hard paste that nobody wants to use. But the OEM has to be able and willing to do this.