Preventing a run out: the right paste makes the difference
In this case, we need a paste that not only has outstanding thermal conductivity (there are quite a few), but also the right consistency (there are already fewer). The paste must be a little more viscous so that it does not run out the sides, because the gaps are already brutal. Pastes like Alphacool Apex or Subzero, which I currently like to use, or the old Gelid PC Extreme are suitable here. Liquid pastes, such as Arctic MX-2 or MX-4, have no place on such a graphics card.
The paste should be at least hand-warm, i.e. between 30 and 40 °C. To achieve this comfortably, you can of course leave it in your pocket for a few minutes and carry it around with you. Below 20 °C, it is somewhat more difficult to process. So the GPU should not be ice cold either. I put the PCB AND the cooler under an infrared radiator after applying the paste, but you can also do this with a hair dryer. But then the paste should be applied afterwards, so that even their surface is not formed skin.
Only for masons: painting with heat conductive precious plaster
I can already spoil it: It is by far the worst method and there is a solid reason for that. On the one hand, you have the bump in the middle and the edges are sloping. The amount of paste applied is not the problem here, but the large-area application is. With a slightly more viscous paste, it will not be possible to push everything out and push it away, because the greater the contact pressure during screwing, the more the paste also accumulates at the edge due to compaction. This becomes a barrier and the paste, which is actually superfluous, is not pressed out completely.
You don’t have this problem with thin paste, but then the paste quickly runs out of the large gaps. Or it dries out and the layer becomes thinner and thinner after thermal heating and eventually cracks at the sides. That is exactly why the industrially used material is hard and strong like concrete. Reliability.
For small chips not even bad: the blob
The blob method is further ahead here, but it is not yet optimal for the large and, above all, rectangular chips. The reason is the required amount of thermal paste, which builds up quite high in very little space. On rather small and square chips, you can happily do it this way, but the large chips are less suitable there. It does not grout well enough and especially not evenly toward the longer edges.
True to the line: It’s all about the sausage!
So, now we come to the real highlight of all my experiments. This method can also be found in modified or combined form on some cooler bottoms with heat-conducting paste already applied. There, you will find the applied thermal paste in strips or honeycombs, but always spaced apart so that it can still spread without bunching up to the point that it prevents the excess paste from escaping. In our case, however, a simple sausage in the middle will do as well if we follow a very specific sequence when screwing it together afterwards. This is elementary and absolutely important!
Let’s therefore now move directly to the screwing. We put the (best) preheated parts on top of each other. If you’re afraid that you won’t hit the hole exactly and then have to push to correct (very counterproductive!), it’s better to use toothpicks as a guide. Simply position the board over the cooler (looking through one of the holes), then hold it in place and carefully insert the toothpicks from above. Then simply set the board down vertically using the guide. Ready! If you have never done this before: Just practice without paste first. Then you quickly have confidence that it will work out well!
But that is not all. Now it comes down to the correct sequence. First, we take a clean, soft cloth and wrap it around the finger. Then, in the middle of the white line, we lightly press the board, but be careful not to slip. Light means easy. Then we alternately screw the two long sides. For example, let’s start on the left at yellow and insert the screws one after the other, turning each until the thread grips. Now we repeat the whole thing on the opposite, light blue side.
Then we alternately turn in the two screws of the long sides a little at a time until we feel the resistance. Then tighten finally. That’s about it. The method with longitudinal edges has been adjusted by some board partners in R&D and has now also been adopted in production. It’s just better than diagonally across. This is only possible with blobs and solid areas. And now it gets exciting, because I measure the whole thing of course!
163 Antworten
Kommentar
Lade neue Kommentare
Mitglied
Mitglied
Veteran
Urgestein
1
Urgestein
Veteran
1
Urgestein
Mitglied
Urgestein
Mitglied
Urgestein
Veteran
Veteran
Veteran
Mitglied
Alle Kommentare lesen unter igor´sLAB Community →