Actually, one should primarily cool the parts directly where the waste heat is also generated. Of course, the shunts cannot be cooled in this way, but coils and capacitors could be connected thermally to a cooling surface. Could, because one has neither made it nor even provided a possibility for it! Because the four GPU waterblocks for the GeForce RTX 4090 FE that I have tested so far are ultimately without any cooling options there as well. The copper block is simply too short on all of them. Incomprehensible, but unfortunately that’s the way it is. This almost seems as if NVIDIA’s Thermal Map never existed.
However, we have to say that NVIDIA itself did not think that far ahead, so the spot in front remains really hot for now. But there are always the backplates, which can be used for stabilization and passive cooling. Does anyone remember my pad mod for the GeForce RTX 3080 FE, which was then quietly adopted 1:1 by NVIDIA into current production? Unfortunately, there are always things that you have to push as an outsider first. Back then it was the hot GDDR6X memory, today it’s just the power connector. There’s always something, it seems.
So I grabbed the aluminum backplate for the Heatkiller V Pro and drilled small 5 mm holes in the relevant places so that I can measure down to the board with the camera. With a high-resolution industrial camera and the 640 measuring points of the bolometer in width, you can even do this without hesitation; with a smaller handheld FLIR or a CAT smartphone, something like this goes terrifically wrong. Nevertheless, I first made a control measurement without placed pads, because I also changed the measured land. The temperatures were absolutely congruent
After that, I placed a thin 1-mm pad in the hollow with the openings to become flat to the rest of the environment. Then I put on thicker, very soft pads (2.5 mm would be ideal, ultra-soft with 3 mm also works) and I carefully cut out the pads with a scalpel at the relevant measuring points. We can see the glowing plug contact with the shunt seat heater and the soldering eyes of the 12VHPWR connection. Here in the picture, for example, I first used the original pad from Watercool, which was actually intended for the GPU. But you really don’t need to cool them from behind.
With this, the pad mod is ready and we can see that it is a very simple solution. We will see how well it performs on the next page with the measurement.
71 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Mitglied
Urgestein
Urgestein
1
Urgestein
Veteran
Urgestein
Veteran
Urgestein
Urgestein
Veteran
1
Urgestein
Veteran
1
Urgestein
Urgestein
Urgestein
Alle Kommentare lesen unter igor´sLAB Community →