Heating and thermal efficiency
To test the heating of the machine, we opted for two distinct tests, first the heating in IDLE and then the heating under CPU load on Cinebench R23 for 10 minutes.
The cooling system
On the outside, we have two openings on the sides of the device to draw in fresh air for the fan. The latter expels the hot air at the back of the mini PC through a third vent.
Inside, we find the fan, blowing air through a radiator connected by heat pipes to the CPU. This system is quite good for the CPU used, but it is still noisy. Indeed, this NUC tends to start the ventilation even when the CPU is below 50°C, which is very disturbing in the long run. The application of the thermal paste is quite well done, but we noticed that the quantity is too large. As a result, the paste was changed and further tests were performed after all our tests were completed.
CPU heating alone
After 8 min of Cinebench R23:
We start with the heating in IDLE and on Cinebench R23 after 10 min of stress. The CPU went up to 100°C on 1 core and 90°C on the others during the first minutes before stabilizing around 80°C. The power consumption and frequency went from 64 W and 3900 MHz to 40 W and 3200 MHz during this stress test. On the surface, we note a maximum temperature on the back of the device with 44,3 °C, the desktop at the back is 32,7°C.
And here is a FLIR image of the top of the device without the cover. As you can see, the hot spot seen previously is in the area of the power socket. We also notice a rather high internal temperature with 47.1°C under the fan.
Disassembly and change of thermal paste
To go further, we decided to change the thermal paste with be quiet! DC2 after disassembly. This high performance thermal paste is particularly fluid and will be perfect for future tests.
So we changed the thermal paste and reassembled the motherboard having first made measurements in the base configuration. Then, three tests on Cinebench R23 were carried out by noting the temperatures at 64 W and 40 W as before. Each of these tests corresponds to a cooling configuration, i.e. closed cover, open cover and motherboard only.
As we can see on the graph below, the gain between with and without cover is about 2°C at 40W with the basic thermal paste. For the DC2 we have a gain of 4 to 5 °C for the measurements at 40 and 64W between closed and open cover. The motherboard alone gains another 3 to 5 °C at 64 W and 1 to 2 °C at 40 W. The most interesting thing is that the maximum temperature went from 100°C open with the basic thermal paste to 87°C with the DC2, i.e. a 13°C gain here. And even more on the motherboard alone with a 16°C difference.
With this gain in temperature, the NUC is cooler in IDLE and under load, and quieter. Even if the ventilation management is still poorly done. This gain in temperature also leads to a gain in surface temperature, with a gain of 1°C on the top of the device and on the desk at the back.