I had recently tested a current and originally packaged Noctua NT-H2 tube, which was purchased anonymously for me by the community and also rated in an article. This paste performed rather mediocre in the test. Due to many requests, the desired follow-up test is now available with an original paste from 2023 (thanks for sending it!) and two other tubes that I subsequently bought myself. I had already written to Noctua after the first measured difference between two samples, sent all the measurement documents and analyses and invested a lot of time in the follow-up, because I am very interested in the topic. So let’s test it again.
And so today there is a transparent post-test and a minor correction in the database, which I am making despite the mixed results. I can already spoil the fact that the better tubes stand out slightly from the first tube I tested, but the paste remains mediocre. I can’t make the effort for every paste to be tested, because the customer doesn’t buy 4 tubes to pick out the best paste, but expects consistent quality. So far I haven’t had any problems like this, but today I’m simply using the current case as an example for a detailed post-test that goes beyond the usual scope because it can provide further interesting insights.
What have I tested now? In the retest, there is the worse paste from 2024 that was previously recorded in the database, a tube from 2023 and two other tubes from 2024. By this I mean my purchase date, because unfortunately the customer cannot see from the packaging what they have actually bought. The filling date and best-before date are clearly missing from the packaging. You could interpret 12.07.25 in the UPC as the best-before date, but that would still not explain when it was bottled. I would expect this from a premium supplier, as well as consistent quality and appropriate performance. In addition to the normal measurements, I also measured the behavior under pressure and the outgassing or BLT increase when the temperature rises between 25°C and 100°C. That will be interesting, I promise.
Reproducibility test with the TIMA5
To safeguard myself against possible doubts, I carried out a plausibility test first this time. After the initial and usual calibration of the system and a smear of the first 4 mm or so, the last tube I bought was measured in three runs. This includes application, measurement and cleaning. The result for the effective thermal resistance of all three runs coincides almost perfectly, although the different position in the tube and any inclusions naturally mean that a certain tolerance limit must be set for such a filling. Under these conditions, however, the result can be regarded as completely congruent, which speaks in favor of the TIMA5 and the reproducibility of the measurements:
Test of serial scattering from four samples
I have now made a complete measurement for all four samples, which has already taken a full day of lab work up to this point. The 2024 sample I criticized (yellow curve) stands out negatively here, the older 10-gram sample from 2023 (blue curve) was slightly more viscous and still deviates somewhat up to approx. 250 µm layer thickness (bondline thickness, BLT), the two other 3.5-gram samples from 2024 (green and red curve) match up pretty well. I therefore decided to consider the paste with the yellow curve as a filling error and to replace it in the database later with the paste with the green curve (sample #3). However, I must emphasize at this point that a difference of almost 0.5 W/m-K in bulk thermal conductivity is nothing to be proud of. However that may have happened. The other three pastes differ within acceptable tolerances.
So today I will compare three pastes in the measurement results of the database. The one from 2024 with the yellow curve, which I consider to be an outlier, the paste from 2023 and one of the two pastes I bought from Amazon a week ago. Then everyone can see for themselves.
Noctua NT-H2 in the first test
Back to the first test: According to its own marketing, the Noctua NT-H2 is an advanced thermal compound based on the older NT-H1. It uses a mixture of metal oxide microparticles, which should offer even lower thermal resistance, thereby optimizing the heat transfer between CPU or GPU and cooler. The paste is said to be characterized by its ease of application, as it does not need to be spread manually. It is not electrically conductive, which minimizes the risk of short circuits, and remains stable over longer periods of time, which should make it particularly suitable for demanding applications such as overclocking or use in quiet systems. That’s what the marketing says, but what does the reality say?
The paste is supplied in a Noctua-typically styled box, together with the tube comes three alcohol cleaning cloths, but that’s it for the very sporty price. Yes, it can certainly be much more expensive, but it can also be much cheaper. And that’s where I see a problem with the outlier. Because you can also assume one thing: it will certainly not have been just one tube. Well, this time the cooler isn’t rattling, but as a tester you should be allowed to question such a scattering.
The fourth paste in the tests is the NT-H1 from 2024 and the DOWSIL TC-5550 as the reference and fifth paste. I will therefore filter the usual charts with actually 10 entries and only deal with these five pastes so that it remains clearer. If you would like to know the basics again and find out more about the test setup and my methodology, you will find the relevant articles behind the two banners:
After so much introduction, we now come to the practical side and the measurements, so please turn the page.
126 Antworten
Kommentar
Lade neue Kommentare
Veteran
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
1
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
1
Urgestein
Urgestein
Urgestein
Veteran
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →