Sign in with
Sign up | Sign in

Thermal Paste Comparison, Part One: Applying Grease And More

Thermal Paste Comparison, Part One: Applying Grease And More
By

If you find yourself fighting a stubbornly-low overclock, there's a chance that your thermal solution isn't working as effectively as it should. We're testing a number of thermal pastes that might help. But first, let's go over the basics of CPU cooling.

I’ve been working on this time-consuming thermal compound test for more than half a year, digging my way through the pastes supplied by Caseking (an online shop in Germany) and the ones we already had on-hand in the lab. Not only does a story like this take a lot of time (it involves nearly 40 products, after all), but it clearly requires a consistent test methodology to make sure the conclusions we draw are sound.

Because we have so many products, we're splitting the story into two parts. The first one delves into the theory and real-world use of thermal compounds, while the second presents all of our benchmark results and the corresponding test setups. 

In part one, we'll cover the thermal properties of CPUs, surface types, background information about various thermal compounds and the methods for applying them, as well as two basic cooler types (liquid and air), along with the issues arising from different mounting pressures. A thermal paste working just fine with one cooler may be a bad fit for another. Therefore, we have to test our thermal pastes on AMD and Intel CPUs with a water cooler, a premium air cooler with high mounting pressure, as well as a more pedestrian push-pin setup, which stands in for the boxed heat sinks you get bundled with most processors.

In addition to CPUs, I also test each paste’s suitability for GPU cooling and assess its viscosity and its ease of use. But let’s start with the basics. What is this primordial goo all about?

The Heat Spreader

When you cut a CPU in half, you notice that the chip (die) itself is much smaller than the CPU package, and thus the die touches only part of the heat spreader. The spreader’s job is to distribute the CPU die’s heat across a larger area, which allows it to pass to the CPU cooler's heat sink.

The drawing illustrates two little-known facts. First, the CPU manufacturer fills the gap between die and heat spreader with a heat-conducting material. While AMD, just like Intel did in the past, fills the void with some kind of solder, Intel now merely uses a thermal compound, which has a higher thermal resistance, but probably saves a few pennies in production. This explains why cooling overclocked Intel CPUs has become more difficult since the Ivy Bridge architecture.

Heat Spreaders, Hot Spots, and Dire Consequences

The drawing also shows that, due to the size difference between CPU die and heat spreader, there are some areas on the heat spreader that will be cooler than the area directly above the die. The area above the die is called the hot spot because it is directly heated by the die underneath. The two images below illustrate what a hot spot is, albeit in an over-simplified way. Reality is not as simple; CPU cores may be loaded differently, and there is also the issue of on-die graphics, which may be more or less active than the processing cores. But let’s just look at the die as a whole and the heat spreader on top of it, viewed from above.

Intel (Core i7-3770K)Intel (Core i7-3770K)AMD (FX-8350)AMD (FX-8350)

Due to its industry-leading 22 nm manufacturing technology, Intel CPUs have a smaller hot spot than AMD CPUs, and you should take this into account when choosing a heat sink. After all, you need to dissipate heat from the hot spot first and foremost.

Benefits and Drawbacks of DHT Coolers

CPU coolers with exposed, ground-flat heat pipes are the latest fad. They certainly save some money during production, which marketing departments then sell to customers as a performance-enhancing feature. But there are drawbacks to this mechanical design. Consider a cooler with, say, four heat pipes, like the Xigmatek Achilles in the picture below. The outermost heat pipes miss the hot spot completely. Even the two innermost heat pipes only partially cover the narrow hot spot of an Ivy Bridge-based CPU. Adding insult to injury, the cooler typically cannot be turned 90 degrees.

The issue with DHT-based designsThe issue with DHT-based designs

If we could turn the heat sink around we'd ameliorate this situation. AMD CPUs are typically not as affected due to their larger die area and CPU orientation; in most cases, all heat pipes cross the rectangular hot spot. If you want a DHT-based cooler, consider one with five heat pipes for more modern Intel CPUs, and try to avoid designs with large gaps between the ground-flat pipes.

Interim Assessment

Just by choosing an poorly-suited cooler, you can lose more thermal performance than the most expensive compound could ever gain back! But there is more bad news. Let’s take a look at what happens between the heat spreader and the heat sink.

Ask a Category Expert

Create a new thread in the Reviews comments forum about this subject

Example: Notebook, Android, SSD hard drive

Display all 136 comments.
This thread is closed for comments
Top Comments
  • 37 Hide
    rolli59 , October 10, 2013 9:29 PM
    Nice article and bookmarked for reference. Looking forward for the next part.
  • 10 Hide
    alidan , October 10, 2013 9:35 PM
    please tell me yo are also going to do the solder the heatsink to the cpu method? i forget what its called, but that is what i want to use for my next computer and would love to see how it stacks up.
Other Comments
  • 37 Hide
    rolli59 , October 10, 2013 9:29 PM
    Nice article and bookmarked for reference. Looking forward for the next part.
  • 10 Hide
    alidan , October 10, 2013 9:35 PM
    please tell me yo are also going to do the solder the heatsink to the cpu method? i forget what its called, but that is what i want to use for my next computer and would love to see how it stacks up.
  • 7 Hide
    The Von Matrices , October 10, 2013 9:36 PM
    In the second section about advanced cooling methods, are you planning on discussing delidding CPUs and replacing thermal paste? If you do it might be worth mentioning that the delidding won't improve temperatures because of improved thermal paste conductivity but because of reducing the thickness of the paste. See http://forums.anandtech.com/showpost.php?p=34053183&postcount=566
  • 3 Hide
    thasan1 , October 10, 2013 9:40 PM
    a really nice and helpful article!
  • 2 Hide
    stickmansam , October 10, 2013 9:44 PM
    Huh, I do turn my heatsinks sometimes for optimal alignment so the heat pipes are perpendicular to the die. Depends if I got the room in the case and what ram is being used. Also heatsink dependent
  • 9 Hide
    slatts1024 , October 10, 2013 9:47 PM
    One of the best articles I've read on Tom's in years and that's saying something. Looking forward to part 2.
  • 6 Hide
    BigMack70 , October 10, 2013 9:48 PM
    ooooooooooh such a tease

    can't wait for part 2 - this was a great read!
  • 2 Hide
    Shankovich , October 10, 2013 9:49 PM
    Loving that DHT-based design overlay picture on the first page. I've been telling my friends for a while to just get coolers with plated covers because the pipes miss the hotspot on intel CPU's, but no I'm full of bs apparently. This video is awesome btw, shows how spreads happen http://www.youtube.com/watch?v=EyXLu1Ms-q4
  • 4 Hide
    nukemaster , October 10, 2013 9:51 PM
    How many volts does this "7 volt" unregulated power supply put out?

    Just curious. I have some 8/9/12 volt regulators that would eliminate the guessing games for resistor fan adapters(voltage depends on the fans current draw).

    I have seen unregulated 6 volt power supplies range from 8-over 12 volts at low loads.

    For a rather low price you can use a regulator to get whatever voltage you want :) 

    ohh yeah and...
    I can't wait for the next part of this to be release
  • 2 Hide
    jimmysmitty , October 10, 2013 9:55 PM
    From what I have seen it depends on the materials. AS5 was great for a while but thee are better ones out than that now such as Noctuas or Zalmans.

    I also enjoyed using the IC Diamond thermal paste as it proved to cool very well but since it has a diamond based substance it can scuff the heat spreader.
  • 6 Hide
    stickmansam , October 10, 2013 9:57 PM
    Also would like to see an Ivy/Haswell test system since they run pretty hot and imo, they need more study on how to best cool them due to the TIM inside them.
  • 2 Hide
    rmpumper , October 10, 2013 10:09 PM
    I always spread the paste manually. Never had issues with overheating.
  • 0 Hide
    smeezekitty , October 10, 2013 10:44 PM
    I am using dirt cheap masscool fanner-420.

    My temps are quite good - really the choice of cooler is much much more important than the choice of paste.
  • 8 Hide
    SteelCity1981 , October 10, 2013 11:07 PM
    "don’t apply too much. Otherwise, the paste will ooze out on all sides. If your paste is electrically conductive, you can almost be assured of hardware damage."

    Tell this to OEM's ever seen a hp, dell, Lenovo, Asus, Acer cpu after you take off the heatsink with thermal paste on it? it's oozed all over he place...
  • 1 Hide
    JimmiG , October 10, 2013 11:24 PM
    I experimented with different application methods when I built my 4770K-based system. Small dot, large dot, X, line. No real difference, but the small dot produced the best results by ~2C. I used MX-4, which is so easy to apply that you can't really mess up.

    When delidding and applying Liquid Pro between the IHS and die, I "painted" a very thin layer. It was very difficult until the surface tension broke, then it was easy from there.
  • 0 Hide
    giovanni86 , October 11, 2013 12:10 AM
    Nice article, you guys ripped me a new one with the spread all over cpu method. Been doing that for a long time hah. Based off what you guys said i feel like it need to reapply paste considering the hot spot on my ivy bridge processor. I feel as if that method works although its very tedious and takes a long time to apply a even surface, i used Arctic MX-2 and it seemed to apply easy with some resistance to spreading evenly flat.. 30 minutes later walaa.. haha painted flat all over.. But i like that pea method in the middle seems to cover the hot spot quite well. Will have to try that out. Can't wait for the next part.
  • 2 Hide
    urimiel , October 11, 2013 12:19 AM
    Dude! this is great. This the first time since the Coppermine era that we started using thermal pastes widely that somebody put all the information together. Congrats guys great job. Well done.
  • 1 Hide
    zyky , October 11, 2013 12:36 AM
    2.66Ghz Q0 Q6600? What planet did I stumble upon? Hopefully one that still has Indigo Xtreme.
  • -9 Hide
    James Hood , October 11, 2013 1:27 AM
    Hey Tom's, I realize that you guys have limited time for these articles, and have to limit the scope. But I think it should be pointed out that not every one reading these articles are complete novices. In fact, I would wager those of us wanting to see the subtle differences between paste and cooler combinations are intermediate to advanced users. As such, in the future I think the risks involved should be left up to the reader. Having the comparison data to go with the risk would be helpful. Gaining .5C better temps for me would not be worth it. But if it dropped 2C on a GPU for my laptop... now that is tempting. As it stands, this data will not be available for those of us that would be interested.

    Otherwise, great article, can't wait to see the results.
  • 6 Hide
    4Ryan6 , October 11, 2013 2:26 AM
    Nice Job Igor, You have your ambient controlled for the tests and very good illustrations of contact imperfections and CPU hot spot area!

    I know you're glad to be nearing the end as that was a lot of time invested to completion.

    Congratulations sir, on one of the best reviews I've seen come from Toms yet!

    Ryan
Display more comments