The issue of applying thermal paste is resurrected in the debate over CPU and graphics card cooling.
A few days ago, Noctua made a quick update on the application of thermal paste to cool processors and, in particular, to anticipate the release of the Ryzen 7000s. Igor’s Lab dwells more on the GPUs of our graphics cards.
Brush, blob or sausage?
Igor Wallossek therefore raised the question of GPU cooling on our graphics cards, noting in passing that the situation is a little different from that of CPUs, given the integration of chips.
For example, the GPU does not take place in a PGA or LGA type socket, and this imposes slightly different constraints. Igor Wallossek has therefore tested several methods of applying thermal paste to come to the conclusion, with supporting photos and measurements, that it is better to make a kind of “sausage” with thermal paste.
The application will then be as uniform as possible, and the temperature could drop by around 3°C compared to an application in the form of a “blob”. This method of application would not be a problem for small chips and is still much better than slathering the entire GPU by spreading the paste all over the surface.
Some “bulk” advice
Spreading the dough would be the worst thing for a GPU to do, if Igor Wallossek’s findings are to be believed. He explains that if you take into account the pressures exerted, it actually prevents the paste from spreading properly all over the chip, which creates areas of “blockage”.
Beyond the application method alone, Igor Wallossek takes advantage of his test to give some additional advice. For example, he very quickly excludes any liquid paste such as Arctic MX-2 or MX-4, and suggests more viscous references such as Alphacool Apex or Subzero.
Igor Wallossek specifies that the paste must be at least 20°C, or even between 30 and 40°C before being applied. This allows for a more even distribution, and it is for this reason that the GPU should not be freezing either. Better, for example, to heat it slightly with a hair dryer beforehand.
To learn more about Igor Wallossek’s testing technique and the results he gets based on his various tests, we suggest you click on the source below. Note, however, that this involves disassembling the graphics card: a manipulation that we will not be – and rightly so – ready to do.
Source : Igor’s Lab
10