Robert Monahan Blog

Replacing Stock Thermal Compound on Two Laptops

This is my attempt at discovering whether or not replacing stock thermal compound is useful endeavor. I undertook this experiment in the spirit of science, and although what follows is not strictly scientific I believe that results could be useful.

Abstract

Replacing the stock thermal compound with a higher quality compound is sometimes toted an effective way of reducing CPU temperatures or increasing performance and I looked to discover if that was true. I carried out three performance tests on two laptops, two stress tests and a test to simulate normal use. All three tests were preformed with the stock thermal compound and a second time with a higher quality Arctic Silver compound. In whole the results show that there is very little difference between the stock compound and high quality compound, there was a slightly lower temperature with the Arctic Silver compound in a test replicating normal use and with one of the laptops a slightly higher frame rate in a stress test. The results show that replacing the thermal compound on a laptop is not especially useful, at best you may see a slight difference on very heavy loads. The results also suggest that heat sink is the bottle neck in the system for heat dissipation.

Introduction

This was an attempt to discover if replacing stock thermal compound with a higher quality compound is an effective way of increasing the heat dissipation from the CPU. Forums on the subject are full of people usually saying one of two things. One; thermal compound replacement will not help much if at all with heat dissipation, and that that stock thermal compound is sufficient. Two; thermal compound replacement is extremely effective at increasing heat dissipation, and the stock compound is insufficient. I even saw some people claiming that thermal compound should be replaced on a regular basis. When looking for numbers on CPU temperatures I saw people claiming no differences, as well as claims saying temperatures dropped as much as 10°C under load. There seemed to be a general consensus that said stock thermal compound effectively as good as any high quality compound and any differences that you might see would be tiny. I suspect that the wilder claims were because conditions on those computers were not carefully controlled, so intended to strictly control an experiment on two laptops to learn if thermal compound replacement is effective modification for a laptop or not. When I started this I was looking for either lower temperatures on equal loads or higher performance at equal temperatures.

Materials & Methods

The experiments were carried out over the course of 3 days on two laptops, a Lenovo Thinkpad X220 and a Macbook 2,1. The X220 has an i7 CPU running KDE Neon 17.10, Coreboot with SeaBIOS and a neutralized Intel ME. The Macbook 2,1 has a core 2 duo CPU running Trisquel 7 with Libreboot, and has holes drilled in the underside for air flow . Both laptops had full batteries, were on AC power, and sat on a desk with normal airflow for all tests. The ambient air temperature was kept at 25°C for all tests and all tests and the laptops were left to cool overnight between the two rounds of tests.

Testing for each laptop was carried out in two rounds, each round consisting of three individual tests. The first round was the baseline test to see how the laptop handled the test with the stock thermal compound. The second round was done after replacing the stock thermal compound with Arctic Silver 5 thermal compound. The three 30 minutes tests that made up the rounds are as follows. First, stress was used to with the command "stress --cpu 2000 --timeout 30m --verbose". Second, a 1080p video of a NASA launch was played on a loop using VLC. And Last, the same stress command and the same video where run together.

The data was collected using psensor and capturing screen shots of the data. Not the most sophisticated way to do it but it was good enough to gain some insight.

Results

Legend

Test 1: stress

Stock Thermal Compound X220

graph of test one, X220 with stock thermal compound

Replacement Thermal Compound X220

graph of test one, X220 with replacement (Artic Silver) thermal compound

There is a slight delay in the the CPU reaching its max temperature for the Arctic Silver test. Also the two core temperatures were much closer to each other with the Arctic Silver which is a trend that continues for all other tests on the X220 model.

Stock Thermal Compound Macbook 2,1

graph of test one, Macbook with stock thermal compound

Replacement Thermal Compound Macbook 2,1

graph of test one, Macbook with replacement (Artic Silver) thermal compound

On the stock thermal compound test the CPU peaks early in the test and then thermal throttles until the CPU is at a steady 83°C. The Arctic Silver test reaches its peak almost about a third of the way into the test and then settles at 85°C

Test 2: video

Stock Thermal Compound X220

graph of test two, X220 with stock thermal compound

Replacement Thermal Compound X220

graph of test two, X220 with replacement (Artic Silver) thermal compound

This test both preformed very similarly except the Arctic Silver ran on average a few degrees cooler than stock.

Stock Thermal Compound Macbook 2,1

graph of test two, Macbook with stock thermal compound

Replacement Thermal Compound Macbook 2,1

graph of test two, Macbook with replacement (Artic Silver) thermal compound

The results again were very similar other than the slightly lower temp on the Arctic Silver.

Test 3: stress & video

Stock Thermal Compound X220

graph of test three, X220 with stock thermal compound

Replacement Thermal Compound X220

graph of test three, X220 with replacement (Artic Silver) thermal compound

Result were very similar on this test. Video playback was smooth and I didn't notice any skipped frames even with the heavy load.

Stock Thermal Compound Macbook 2,1

graph of test three, Macbook with stock thermal compound

Replacement Thermal Compound Macbook 2,1

graph of test three, Macbook with replacement (Artic Silver) thermal compound

The stock ran at about 81°C once the temp settled, and the frame rate on the video was aprox. 1-3 frames/second. The Arctic Silver test ran at a warmer 84°C and the video playback was about 3-10 frames/second.

Discussion


Explaining the results

The Arctic Silver compound seemed to have a higher thermal conductivity than the stock compound and that is most likely why the Arctic Silver tests seem to take a little longer to reach the temperature plateau and the temperatures didn't spike as much. Once the heat sink is at thermal capacity the differences between the compounds becomes negligible. It seems that the heat bottle neck is at the heat sink to air transfer and not at the CPU to heat sink transfer.


Is replacing thermal compound useful

The results were interesting and I believe can shed some light on the usefulness replacing the thermal compound on a laptop. Over all it does not seem that replacing the thermal compound is an effective way to see large improvements in performance, but it may allow you to eke out some small improvements on heavy loads. As a general rule I would not recommend replacing the thermal compound for three reasons. First, it was a bit of a pain to open and replace thermal compound and for average users there is a real risk of damaging something or misapplying the compound resulting in higher temperatures. Second, the improvements in performance are tiny, and when using the laptops for normal daily tasks there was no noticeable improvements. Third, there are easier and more effective ways of increasing the performance such as using a external fan or A/C unit to cool the laptop or using less CPU and GPU intensive applications.


What I would do differently

If I were to test thermal compounds again there are a number of improvements I would make in my methods. First and most importantly my method of data collection (i.e. using psensor and taking screen shots) was bad, I wasn't able to easily analyze the data, create trend lines, graph lines together, etc. It would have been much better to get the data and record it in a spreadsheet so that I could build my own graphs and to extract the most important information. Second, I only collected CPU temperature and fan speed data and it would have been interesting to collect other bits of info such as the amount of work that was completed by stress and the actual FPS of the video playback. Third, I would have liked to test a few other computers to see if the results are the same and I'd like to test a desktop PC with a large heat sink and good air flow.