Warning: Always look at the date when you read a hardware article. Some of the content in this article is most likely out of date, as it was written on November 12, 2012. Check out our more recent articles. | |||||||||||||||||||||||||||
IntroductionNo man-made surface is perfectly smooth, and the tiny gaps between a CPU and a heatsink are usually large enough to have a negative impact on cooling. To combat this, thermal paste is used to fill in these air gaps. The thermal paste itself is not as thermally conductive as metal, however, so it is important that not too much is used. If that happens, the thermal paste will behave more like an insulator than a means of improving thermal performance. The best technique to apply thermal paste is something that is often debated on the internet, so in this article we want to do our own testing to find the best method for applying thermal paste. First we will see how each technique spreads the thermal paste across the CPU, and then we will test each technique in a full computer system to see which has the best thermal performance. | |||||||||||||||||||||||||||
Test SetupTo get a clear view of how the thermal past spreads for each technique, and to see if the spread results in any air bubbles, we fabricated a 3/8" thick acrylic plate to act as an interim heatsink. This acrylic plate has potions removed to clear the capacitors around the CPU, and the mounting is the exact same size and spacing as the Corsair H60 CPU Cooler. This way, we can ensure that our results match exactly what you would see when using a real heatsink.
We are using the built in video controller to help put just a little extra load on the CPU which should accentuate any temperature variances. An open-air test platform was used in place of a chassis since it should allow for more consistent results. | |||||||||||||||||||||||||||
Application TechniquesThe techniques we are testing come from the common recommended techniques found online as well as a few we devised ourselves. We made sure to apply the paste evenly, but did not make it 100% perfect. We did this on purpose so that our results would be consistent with what you would see in the real world.
| |||||||||||||||||||||||||||
Spread ResultsIt is almost impossible to see the tiny air bubbles in the images below, so we went ahead and circled them in blue. Larger air bubbles have larger circles, while smaller bubbles have smaller circles. Similarly, any areas where the thermal paste did not cover the CPU are circled in red.
Both roughly and smoothly spreading the thermal paste across the CPU had great coverage, but resulted in quite a few small air bubbles. These air bubbles are tiny compared to the circle and spiral shapes and likely to have a minimal effect on performance, but they are definitely present. Spreading the thermal paste perfectly should result in fewer air bubbles, but in our experience it is impossible to get it 100% perfect. Surprisingly, the X shape resulted in the best overall spread with great coverage and very few air bubbles. This surprised us a bit, but makes perfect sense when you think about it. The X shape allows the paste to spread equally across the CPU, and since the spread is from the inside out, the shape helps prevent air pockets from becoming trapped. Unfortunately, the happy face did not have great coverage and resulted in a few relatively large air bubbles. But the biggest factor for each technique is how well each actually cools the CPU, so let's take a look at our thermal testing results: | |||||||||||||||||||||||||||
Temperature ResultsOverall, the thermal results are pretty close to what we expected given our spread testing. The X shape, with the best coverage and fewest air bubbles, performed the best at 54.25°C under full load. The smooth spread was close behind and was only .25 °C hotter under full load. Surprisingly, the rice sized dot tied with both the happy face and the circle with dot for third place at 54.75 °C. Interestingly, the 2x rice sized dot and thick line - both of which had quite a bit of excess thermal paste that got squeezed out from between the CPU and the heatsink - did the worst in the thermal testing. This reinforces the idea that, just like too little thermal paste, too much thermal paste can result in sub-standard performance. | |||||||||||||||||||||||||||
Conclusion
Perhaps our most interesting result is that if you are going to use the tried-and-true rice sized dot, you might as well have a bit of fun and draw a little happy face instead. Not only does it cool just as well, but you can rest easy knowing that you have a happy, smiling CPU underneath your heatsink.
| |||||||||||||||||||||||||||

What would be the most optimal amount of paste and spread method for a processor without an integrated heatsink? For instance, a laptop processor with just the die coming into contact with the base of the heatsink?
For something like that, I would use the smooth spread method. Getting perfect coverage on a non-square shaped die is going to be difficult unless you do the spreading yourself. If you are unsure if the coverage is good or not, it's usually a good idea to apply the thermal paste, install the heatsink, then remove the heatsink and see how well it spread. Check that the entire die is covered and that not too much was pressed out the sides. You will want to clean off the die and heatsink and reapply the thermal paste, but at least you will know for sure if you are using the right amount.
Great test.
But how about a test the other way around. Take different heatsinks and test how they fair with the different metods? Some heatsinks that have the copper tubes directly onto the CPU, have gaps bethween the tubes, which metod is right for that
In other words, as long you use the right amount, any of the tecniques are the same realy. So its up to what one the person has more experience and confort to apply.
I think the reason your smooth spread has so much air trapped was because it was not smooth. It had a lot of texture to it. I just installed a cpu today using the smooth method and I took the time to smooth it as flat as possible. The result was a CPU that with the stock cooler was easy to OC about 300 mhz above stock (3.8ghz to 4.1 Ghz) speed and still be within safe operating tolerances.
how to apply the or spread the paste ?? plz elaborate
what did you guys use to torque the heat sink dummy down with? or did the screws have shoulders which allowed them to only tighten to a certain clearance? Also it seemed like a lot of the other shapes could have worked better if they were just applied thicker and had more squeeze out.
Shane
This is what Cisco recommends for use on Xeon CPUs in some of their blade servers, so it looks like the X shape is well-regarded.