Can a thermal paste scratch your CPU die?

The actual area on the desktop is not increased, if the die is the same then the heat spreader will not help contact, the difference comes from the sheer size of the heatsinks.

The heat spreader actually hurts the thermal conductivity by adding another layer which is why the serious tweakers will de-lid their CPU and run a manually adjusted heatsink to apply the correct pressure.

 

Agree there is an added a couple of layers of resistance but the IHS does spread the heat over a larger area so your heat transfer is spread out. And added resistances are compensated for with the larger sinks as you said.

The restricted cooling in laptops will increase the heat density of the chip over a desktop hence the higher Delta T's

 

Those are some terribad heatsink contact indeed. But I'm hoping those are either outliers, or contain ULV chips that don't give off too much heat. or maybe you just used pressure film with too high of a sensing threshold

But seriously, this is what a good heatsink contact should look like:


Sensing threshold = 28 PSI. As you can see it's still not perfect, especially at the two corners, but the coverage is more than adequate, and in this case less TIM actually is more.

 

Where do you get that contact paper to test?

 

Actually same paper 28-90 PSI ULL

Here are some of the samples taken from this forum just scroll down

Reliability

Only one of the eight has what I would call good contact- 7 were in the 10-30% contact range, pressure is pretty light too.

In manufacturing production I would be surprised if they could hold a tolerance of +/- .005 on multiple twists and bends, pressing, machining etc. to get flat and parallel mating surfaces.

Just the nature of the beast.

Your result seems atypical, better than most- I always thought there might be an opportunity in a custom fitting service for notebook sinks, provide a certified contact and pressure via paper and contact lab report. Temperature would drop probably 10C and compound life expectancy would increase 2-3X or more as poor C/P will drastically reduce performance and reliability.

 

PM me your address and I will send you some

 

You can request a free sample from SPI

Unfortunately the market is just far too small for such a niche service, it would have to be a side business for it to be viable.

Also, despite the good contact, switching from MX-4 to Liquid Ultra still provided a 7-10C drop in temperature under load. I honestly think this is due to the excellent wetting properties of Liquid Ultra and how thin it can be spread rather than due to its thermal conductivity. And also goes to show how filling in all those microscopic gaps can yield results even when contact appears to be good.

 

Air happens to have one of the worse thermal conductivity, so filling microscopic gaps usually filled with air will indeed make quit the difference if there are enough of those.