Burn-in measured: Honeywell PTM7950 Phase Transition Material examinated and how to recognise a copycat

1 - Introduction and necessary definition of terms 2 - Measurement of the Honeywell PTM7950 in detail 3 - Comparison with a another pad (OEM) and conclusion

The burn-in of the Honeywell PTM7950 in detail

Measurements were taken with a force of 8.5 N on 100 mm², which corresponds pretty much exactly to the recommended value of a standard GeForce RTX 4080 and between 25 and 75 °C for the average pad temperature (including the gradients) in steps of 5 degrees each. Let’s now start with the very lengthy process, which I performed a total of 10 times. The diagram shows the first burn-in using the yellow curve (Rth, Y-axis), the blue dotted curve shows the resulting BLT in the temperature curve (X-axis) and the red curve shows the Rth at the 10th run. However, the layer thickness remained constant at approx. 16 µm from the second run onwards (< 40 °C approx. 17 to 18 µm). We can see that the PTM7950 exhibits an almost linear decrease in BLT up to approx. 45 °C and that the thermal resistance also decreases accordingly up to approx. 40 °C and then remains constant from around 50 °C.

So you can see that you should actually reach 50 °C in order to complete the burn-in, whereby the temperature window is between 40 and 50 °C and Honeywell specifies the average temperature here. This in turn refers to the median temperature of the pad. However, I never claimed that the pad would not work after the burn-in below this temperature limit, because it would not make sense as a statement (see red curve). Yes, the Rth is still slightly higher, but this difference can certainly be completely neglected in real life.

Incidentally, I am working here with a threshold of 0.25 K/min as well as a window of 100 measurements and a waiting time of 2 seconds, during which all conditions within this frame must be fulfilled. This is certainly much faster on a PC, but I wanted to record it as accurately as possible. Speaking of the PC, we also have the graphics card. Let’s just compare what causes the burn-in here and how to recognize it.

With the air-cooled graphics card, GPU Edge already needs almost 69 °C to convert the entire pad once, but I also have an interesting graphic that shows the burn-in (with a slight time delay) from the perspective of the air-cooled GPU. We can see that the burn-in takes a good 5 minutes here, as I couldn’t and didn’t want to simply switch off the fans. Due to the lower interface resistance and the thinner layer of the melted pad alone, the GPU edge temperature drops by a whopping 3 Kelvin (with the hotspot it’s even over 5)!

Now we come to the point where I have to correct myself in parts, because the temperature for the burn-in with the PTM7950 was significantly lower than with many conventional (older) PTMs and the window was also significantly wider. If you know the delta between the GPU temperature and the water of a GPU water block, you will be able to estimate quite quickly, even without the TIMA, what temperature you really need in the GPU for a safe and complete burn-in in a specific setup. I have simply measured both sides, i.e. once the top side towards the heat source and once the side towards the cooler, to illustrate the delta within the pad layer.

I wrote in the product test of the PTM7950 that I see problems with water coolers and water blocks where a complete phase change could not occur, at least on the cooler side, and that you would need at least 55 to 60 degrees in the entire pad for a complete change of state. Depending on the pad, it should be possible to correct this temperature downwards. As the phase change is completed at 50 °C for the PTM7950, you should expect 52 to 53 °C for the entire pad to change to the new state.

That brings us to the Honeywell PTM7950, but not to the other questions that have reached me. This also includes the desired comparison with OEM pads that are offered as PTM7950 but are only 0.2 mm thick. I have one of them here and I have also found out what it might actually be. But you’ll have to turn the page one last time.