Linus Tech Tips PTM7950 Review – Original, OEM Product or Fake?

1 - Introduction and overview 2 - Microscopy, pad thickness and homogeneity 3 - Microscopy, particle analysis and chemical composition 4 - Burn-in and performance measurements 5 - Summary and conclusion

Burn-in behavior and melting of the pad

During burn-in, the LTT pad exhibits behavior that is almost identical to the well-known Honeywell PTM7950 in many respects. Under the defined load of 9 Newtons, the bond line thickness of both PTM7950 samples begins to drop significantly at around 35 degrees Celsius. It is precisely in this area that the actual melting process of the polymer matrix begins, the material becomes softer, starts to flow and the initially much thicker layer is successively compressed under pressure. The Laird TPCM 7000 and comparable OEM pads react much more slowly in this temperature range and only show the characteristic, steeper drop from around 40 to 45 degrees Celsius. They therefore require a noticeably higher temperature until the phase change or the relevant softening of the matrix occurs.

The end point of the burn-in for all three pads is around 55 degrees Celsius, from which point the BLT curves are almost horizontal and the layer thickness changes only marginally. It is interesting to note that both PTM7950 pads, i.e. the reference material from Honeywell with a nominal 0.2 millimeters and the LTT pad with a specified 0.25 millimeters, can ultimately be pressed together to practically identical values at the applied 9 Newtons. The measured bond line thickness ended up being exactly 14 micrometers. This indicates that the effective compressibility and the filler structure of both PTM variants are very similar and that the matrix transforms into a comparable, densely packed final structure once the phase change is complete.

However, there is a noticeable difference in the shape of the curves. The original Honeywell PTM7950 shows a clearly pronounced kink in the area around 45 degrees Celsius (exactly as shown in the Honeywell data sheet), i.e. a peak in the course of the BLT reduction before the layer collapses further and stabilizes at higher temperatures. This kink is typically associated with the melting of a certain fraction of the polymer phase, for example due to different melting points of several paraffinic components or a defined crystallite structure that collapses in a narrow temperature window. The LTT pad lacks precisely this distinctive peak; the BLT curve is much smoother and looks more like a continuous transition without a pronounced intermediate stage.

There are several possible explanations for this, which I can discuss on the basis of the available data without being able to verify them conclusively. One possibility is a slightly different distribution of the paraffinic components in the polymer matrix, such as a narrower melting interval or a slightly different crystallinity, whereby the phase change does not take place in a sharply defined temperature window but is more widely distributed. A somewhat stronger pre-compaction or a slightly different initial state of the layer, which smoothes the mechanical reaction, would also be conceivable. However, the microscopy and the LIBS analyses indicate an almost identical filler combination of aluminum oxide and zinc oxide as well as a silicone-free polymer matrix, so that fundamental differences in the material design seem rather unlikely. According to the cross-section measurements, the actual initial thickness of the LTT pad is also closer to 0.2 than 0.25 millimeters, which can also influence the process because a slightly thinner layer has to collapse less in order to achieve the final bond line thickness. Ultimately, it should be noted that although the LTT pad shows the same melting start, the same end point and the same final thickness, the characteristic intermediate kink of the reference pad is significantly weakened or practically smoothed, which indicates a slightly modified thermomechanical reaction of the polymer phase.

Temperatures at the top and bottom during the melting process

Measuring the surface temperatures during burn-in provides additional information on the phase change and the behavior of the coating. The curve of the “hot side” increases with the applied sample temperature, the “cold side” follows with a slightly lower curve. The increasing temperature delta between both sides of the pad clearly shows how the thermal resistance changes during the melting process. At the beginning, at temperatures in the range of 25 to around 35 degrees Celsius, both temperatures are still relatively close to each other, the material is largely solid, the effective heat conduction distance is large and the internal reorganization is only just beginning.

As soon as the phase change begins, the gap between the warmer and colder sides opens up more clearly. In this phase, the structure in the layer is continuously remodeled, air spaces or cavities are reduced, the particles shift towards each other and the polymer temporarily loses some of its mechanical stiffness. The diagram shows that the hot-side temperature increases slightly disproportionately, while the cold side lags slightly behind, which increases the delta. This is physically plausible because the resulting, not yet fully consolidated structure initially has a higher local thermal resistance. As the temperature increases and burn-in progresses, the system then approaches a new equilibrium in which the layer thickness is reduced, the particles are packed more densely and the interfaces are wetted. In this state, the temperature delta stabilizes again at a smaller, almost linear value.

The fact that this transition in the LTT pad is similar to that in the Honeywell reference pad supports the assumption that the phase change mechanism is basically the same, even if the kink in the BLT curve is less pronounced. The growing delta and its subsequent stabilization are typical features of a functioning PTM burn-in process, which leads from an abruptly changing mixed state to a homogenized, thermally more efficient layer.

Thermal resistance in comparison and interpretation

The proximity of the materials becomes even clearer in the diagram of the thermal resistance over the burn-in. At the beginning of the test, the LTT pad starts with the highest Rth value in the field; it is clearly above the Honeywell reference pad and also above the Laird TPCM 7000. This is not surprising, because the initially greater effective layer thickness, the higher internal porosity and the still insufficiently consolidated interfaces initially maximize the thermal resistance. However, as the temperature increases, the Rth drops very rapidly, falling below both the Laird pad and the Honeywell PTM7950 in succession from around 40 degrees Celsius. At the end point of the burn-in, i.e. in the range of 55 to 60 degrees, the LTT pad has the lowest thermal resistance of the three candidates.

There are several possible, mutually complementary explanations for this. Firstly, the strong compression under 9 Newtons compresses both PTM pads to an identical 14 micrometers, while the Laird pad retains a greater residual thickness due to its slightly different matrix and compressibility. As the thermal resistance is essentially proportional to the layer thickness with the same thermal conductivity, a difference of just a few micrometers is sufficient to produce measurable differences in the Rth. On the other hand, minimally better wetting of the interfaces can play a role with the LTT pad. If the polymer phase in the burn-in is somewhat more easily incorporated into the microstructures of the chip and cooler, the additional interface resistance decreases, even if the bulk conductivity of the filler matrix is identical. The microscopy showed a very homogeneous filler distribution without conspicuous agglomerates or binder-rich zones, which is consistent with a uniform, easily packable final structure.

Based purely on the Rth curves, no fundamental deviation from the PTM7950 behavior can be seen; on the contrary, the LTT pad not only ranks in line, but is even slightly ahead of both comparison materials in the end. Whether this is due to a slightly more favorable real final thickness, minimal differences in the filler volume fraction or optimized wetting, I cannot clearly verify with the available data. However, the overall trend is clear: after complete burn-in, the LTT pad offers at least the same thermal performance as a good PTM7950, with a slight advantage in thermal resistance under the test conditions selected here.

GPU temperatures, practical relevance and overall conclusion

The technically accurate GPU temperature simulation using a GeForce RTX 5090 with a 600 watt load confirms this picture in practical use. At the start of burn-in, the GPU temperature with the LTT pad is the highest because the layer is still thick and the thermal resistance is correspondingly high. After just a few cycles, however, the curves of the LTT-PTM, Honeywell-PTM7950 and Laird TPCM 7000 come very close together. In the further course and especially in the steady state after completion of the burn-in, the GPU temperature with the LTT pad is slightly below that of the Honeywell reference pad and also slightly below that of the Laird pad. The differences are in a range of well below one Kelvin, i.e. small, but reproducible in direct comparison and consistent with the previously measured Rth differences.

From a thermal point of view, this indicates that the LTT pad has the lowest total resistance in the cooler-pad-GPU chain after correct burn-in. The impact on the absolute GPU temperature is of course limited given the overall resistance chain, but it matches exactly what would be inferred from the ASTM measurements of BLT and thermal resistance. Microscopy of the surface shows a very fine, homogeneous filler structure without noticeable inhomogeneities, the 3D profile measurements of the layer thickness show a real raw thickness that is more in the 0.2 millimeter class, and the LIBS analysis shows an almost classic PTM7950 signature with aluminum oxide and zinc oxide as the main fillers and a completely silicon-free polymer matrix.

The sum of the observations thus gives a consistent picture. The chemical composition, the filler morphology, the particle size distribution, the burn-in behavior with early melting start at 35 degrees Celsius, the end point at 55 degrees, the identical final bond line thickness of 14 micrometers and the low thermal resistance in full operation all indicate that the pad offered in the LTT store is functionally and structurally very closely based on the well-known PTM7950. The lack of a pronounced kink at 45 degrees Celsius indicates a slightly smoothed or minimally modified thermomechanical reaction of the polymer phase without changing the basic characteristics. Although I cannot derive an official declaration of identity from the available measurements and thus cannot verify a one hundred percent similarity, everything that can be measured and microscopically recorded clearly places the LTT pad in the same material family and explains why it behaves thermally on the GPU in the same way as one would expect from a very good PTM7950.