Measurement of board temperatures in the transition to the terminal
But how much do the circuit board and the contacts heat up now? I now measure the back of the graphics card and use the original 4-way adapter from NVIDIA, where I only removed the summary of the wires and the sleeves, because I did not only want to measure with the IR camera, but also used a temperature sensor that directly monitors one of the 12V contacts. The board is still coated with a calibrated medium at the relevant points: transparent, silk-matt measuring lacquer, as used for the “tropicalization” of boards.
I’ll first show you the first 45 seconds after applying the load of 600 watts in real time and you can see that the area in question heats up just as fast as the water-cooled voltage converter area for NVVDD (GPU voltage). The measuring point “PCB Below Contacts” is located on a soldering eye of the 12VHPWR connection and also represents the temperatures of the plug contact (Long-Term) with a delta of maximum 2 Kelvin. After all, we mustn’t forget one thing: The connection is heated up considerably from exactly this point.
You can now see the entire 30 minutes of the measurement process shortened to 30 seconds in fast motion, because otherwise you would surely fall asleep in between. From minute 10 on, relatively little happens anyway, because everything is already almost completely heated up.
Afterwards I made a super projection, where you can also “see” the board from the front side at the right position. The hotspot is with its 59 to 60 °C below a shunt on the front side and seriously, which really strongly optimistic designer puts such a firefly (in real over 100 °C) directly above a mini plug contact on the back side of the PCB? But that’s another story. The hotspot alternates between shunt back and a contact of the left of the two input coils. The solder eye on the 12VHPWR reports plenty of 57 °C and then inside the connector, the contacts are back to about 60 to 61 °C.
However, it is a fact that almost all of the connector’s waste heat travels from the graphics board to the connector and is NOT a result of the contacts heating up due to excessive contact resistance! On the contrary, the colder the board becomes, the lower the proportion of contact heating. At 450 watts, it is again less than 2 Kelvin (temperature difference, delta) that are added on top. Nothing more. So we see that the uncooled board is responsible for almost 94% of the contact temperature.
And our test object is water-cooled! In the air-cooled setup, I had already measured around 82 to 84 °C on the board, with the contacts then going towards 88 °C and more. For security reasons, I aborted this after about 15 minutes, because I only have this one card. But it should not be left unmentioned. And now? Tinkering again, what else!
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 →