Diamond Thermal Paste: Update

thermal_paste_test_rig

The need to conduct laboratory-style experiments runs deep in some people. [Freddyman] built an apparatus to test out several commercial and homemade thermal pastes, including the DIY diamond thermal grease we reported on last month. He setup each experiment in the middle of an air conditioned room, ran the heat sink fan for 30 minutes to equalize the temperature, then turned on the DIY heat generator that the paste and heat sink were connected to. He’s got a lot of data from tests he ran with the eight thermal conductors; air (using no paste), Arctic Silver 5, Ceramique, Dow thermal fluid, pure silicone oil, silicone and diamond slurry, Dow fluid with diamonds, and the Inventgeek.com remake.

One of the big problems with DIY paste is the air bubbles that are introduced into the slurry as you mix in the diamonds. All of the homemade pastes except one were put in a vacuum chamber in an attempt to remove tiny bubbles. The one that wasn’t put in the vacuum performed the worst of all the thermal conductors. In all cases, the commercially available products performed quite well while the DIY solutions delivered mixed results.

18 thoughts on “Diamond Thermal Paste: Update

  1. Great job.

    It’s nice to see people going out an doing comparisons of things that people wouldn’t normally compare.

    I mean how many people (except us on hack a day) would even think twice about a comparison of thermal pastes.

  2. I am actually using the inventgeek thermal compound on my Intel Quad Core right now with very good results. i haven’t testes the other compounds but it is worth mentioning this is the first DIY Thermal compound ever. and while the results of testing may be flawed with his methodology, he as inspired alot of people to experiment and opened this up as world to experiment in!

  3. I’ve said this before and I’ll say it again: you cannot test Arctic Silver 5 seriously until you have let the paste cast for the specified amount of time. It’s something like 100 or 200 hours with periodic cycling between hot and cold. Usually temperatures drop about 2-3 C after this is done.

  4. Ya know… reading that forum and even some of the posts here I can’t help but feel a certain lynch mob attitude is forming. I thought it interesting that hackaday published their original post about that article with no reference that it is 3 years old, and at the time there was nothing to compare it to on the market. Basically I’m starting to feel hackaday is somewhat responsible for a very negative attitude about an author that has had a tremendous amount to contribute to hackaday and an even greater involvement in the comments defending hackaday. Kind of feels like hackaday is hanging him out to dry out of context to me. Where’s Elliot now?

  5. an interisting and applicable quote from the forums…

    “… your results are just as flawed as his are. I know of only one truly scientific study that used the appropriate scientific equipment to determine actual results. Overclockers.com conclusively showed the power of diamond based cooling a year before the InventGeek article. InventGeek even reference it. The article has been ironically removed but is available on the way back machine. So before you go throwing the first stone remember your testing methods are equally flawed though well thought out. In the words of the original article by oc.com, “a specially designed and built calorimeter was used. This is a precise instrument that cost about $75,000 to build.”. Secondly I would never test with or use in construction for test equipment wood. It is a tremendous insulator and with each test you change the thermal characteristics of the material with repeated attaching and seating of your heat sink and smearing of thermal compounds absorbed by the wood. Finally you didn’t even test and compare the actual test they did using arctic silver, so you’re testing is far from conclusive. Unless you reproduce his exact environment you are not proving anything. Though kudos for the efforts!”

  6. One possible methodology error is how he mixed the DIY compounds. In the inventgeek article they used some ghetto paperclip in a gum container contraption. In this aritcle he states that he used a dremel with a special mixing attachment.

    Then he comments on air bubbles being a factor.

    Anyone who knows anything about baking would know, one of the main reasons you use a high speed mixer is that it adds air bubbles (thus creating a lighter final product).

    He acknowledges that air bubbles are an issue and then treats his other DIY compounds to a vacuum to remove them… except for the inventgeek compound.

    Not saying the inventgeek compound would have won, but it seems disingenuous to handicap it as much as he did.

    (For those that will say: “But inventgeek never said to use a vacuum on the compound.”, I say: “Yes, but they also didn’t say anything about using a dremel to mix it.”)

  7. I am now actually far more fundamentally disturbed than I thought possible. I just spent the last 2 hours going through every post on hackaday… no small task… and this is the first time hackaday has attacked someone openly. What makes this more grievous in my eyes is the fact that the person attacked is a prolific author and has contributed a great deal to hackaday and in many cases were the only posts of fresh content when the site was young and content starved. If this is how hackaday will be operating now I recommend that no one contribute to a machine that now attacks its fan base and contributor pool. This is absolutely obscene and wrong regardless of the assigning articles and rebuttal articles. We post hacks. We post our experiments. We don’t attack another person for having integrity and follow through in highly experimental areas. Hackaday shame on you today.

  8. you can argue about infogeek goo being better/worse all you want, what you are all missing is the conclusion – difference between best stuff he had and “Pure Silicone Oil” is 5’C at MAX.

    Thermal Conductivity (g/cal/cm/sec °C) = 0.00038
    thats 1.59 W/(m K). This is normal for cheap silicone paste. I use stuff that has 0.78 and I am happy with >3GHz OCs http://www.angela.pl/p4782,ag-chemia-pasta-silikonowa-termoprzewodzaca-h-100g.html

    AS5 has ~9 W/(m K) according to manufacturer, but 0.9 W/(m K) in reality (IBM lab http://www.electronics-cooling.com/html/2009_feb_a2.php )

    There is a lot of marketing and lies surrounding this subject because there is a LOT of money to be made on stupid people buying paste for $20 a pop.
    Conclusion is simple – 5’C difference.

  9. Rasz is correct, with 5 deg difference from the best performer to the worst, that is no real difference at all (maybe to hardcore overclockers only). The average gamer will have their cpu run anywhere from 50-60 deg, what does 5 deg extra mean to such a person? Diddly squat since most cpu’s can run stable in excess of 70 deg!

    Heck I used to use some cheap ass white silicone compound on my old gaming rig, the temps never really got high and the rig is still running fine today. I intend to use the same cheap ass stuff on my newly bought i7 processor (Ebay is king for cheapness lol), given I only intend to play games and no overclocking malarky. I do not need an E-penis to say my cpu can do 5ghz quad core etc, as long as it plays the top end games at max quality, that is all that is needed.

  10. @jayson anders: This is NOT the first DIY thermal compound, and is not the first DIY diamond paste compound. You are giving him far more credit than is due.

    inventgeek posts (in my opinion) clearly fudged data, in an attempt to receive attention. Sadly it works.

  11. Producing a thermal diamond paste is not as easy as the DIYers think it is.

    I happen to be in the industrial diamond field and we have worked with major companies to overcome issues involved. It’s very difficult to have an air free application. Also the thermal properties and efficiency vary with diamond type, size and shape. Dispersion and suspension are also critical.

    Any real success would be truly accidental in an amateur situation. Like anything else, it’s wise to know your materials.

  12. Meanwhile back at the original topic, I have another candidate that should/could be tested.

    I needed some paste the other day for a giant LED array (9000lm !!) and couldn’t find my tiny $10.00 tube I bought a few months ago. Arghh.

    Have you ever seen anti Seize compound? It is powdered metal in a grease base. Maybe it has some metal plating or oxygen excluding chemical in it. It’s used on exhaust manifolds, spark plugs and anything else that may become seized and result in tearing the threads out of the softer metal in an assembly.

    Whoa! It’s thick and sticky just like thermal paste. It is available in copper and nickel and perhaps some others.

    So far It has worked OK, I haven’t fried any big components with it yet. Is anyone interested in testing this stuff?

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.