Real laboratory tests of thermal pastes on igor’sLAB – Part 2 – First comparative measurement with 2 expensive pastes

1 - Thermal resistance instead of thermal conductivity 2 - Practical tests and summary

Yesterday we already had the big theoretical part, today we are doing a practical measurement with two selected pastes, whereby the paste that is nominally better on the data sheet is actually the “worse” one in the test and vice versa. Of course, the factor of potential long-term durability also plays a role, which may put the result into perspective. So let’s take a look. I have named the pastes “Reference” and “Gaming Paste A” because I don’t want to spoil the fun of the individual tests later on. You’ll soon find out which is which. And I’m not using the reference paste as a reference without good reason, as it’s basically an unadulterated industrial paste with guaranteed consistent quality, which I wouldn’t bet on with the usual bottlers.

But today we will see what else I can do and evaluate with the values determined by TIMA5. That’s quite a lot. And the good thing about it: all the measurements are reproducible and you would already recognize blatant errors from the curves of the thermal resistances, whose curves are ideally linear. At least until the paste falls apart if too much pressure is applied.

The effective thermal resistance is the most important factor of all

Let’s start with the most important aspect, the thermal resistance. I already explained this to you in detail yesterday. The most important property is that this correlates nicely linearly with the layer thickness, while the thermal conductivity describes a completely different curve and remains anything but linear. But more on that in a moment.

We are interested in layer thicknesses of 100 µm and below, everything else is really for the gallery. Some manufacturers also state the pure, idealized bulk value here, but this is so unworldly that it makes you want to cry. We can see from the graph that the paste specified in the data sheet with 9.7 W/(m-K) performs better than the paste with the supposed 17 W/(m-K). The thermal resistance of the Gaming Paste A is significantly lower and we will see at the end what this could mean when extrapolated to a CPU with the same heat dissipation.

In the data interface, you can check the determined values again and deselect the deviating values (here everything from 25 µm downwards) for the determination. At this layer thickness, the paste already showed slight signs of dissolution. For whatever reason. The slightly poorer reference paste did not show this behavior. But more on this in a moment.

But at least I wanted to know how far you can go with proper pressure (I could have pressed the whole thing with 300 N, but then nothing changes) and what minimum layer thicknesses can still be achieved. The gaming paste A is significantly “muddier” than the reference, but it has to make do with less heat-conducting small filling particles of zinc oxide. What does less mean, it has none at all. The winner is the polymer pad as a phase change material (PCM), which is truly unique.

The effective thermal conductivity is only an accessory

I already wrote that you can hardly recognize or compare anything on the basis of thermal conductivity, because we only see a few curves here. Anyone who now believes that thermal conductivity is a constant value is very much mistaken. But I already explained this at length in yesterday’s basic article. Yes, everyone is clamoring for it and that’s why I include it in the tests, but if you have Rth, you don’t need λeff_, i.e. the thermal conductivity. And the pure specification for the idealized bulk value even less so.

We will see on the next page whether the difference between the two pastes is really as great as the curves here suggest.