Passiv cooling device for orion

#21

Finally, I took the time to compare the stock Radxa cooler and the bulky aluminium passive 2U heatsink. This is a quick and dirty test, in the sense I just rely on what I have to compare and report, but it should be sufficient to get an idea of what is suitable.

I checked both power and temperature on my Orion board, in closed conditions. The computer case is a cheap mini ITX Aerocool CS-106 case in which two 120mm fans are installed (intake fan in front panel, outlet fan at the back). There is currently no discrete graphics and I use the Radxa image of Debian 12 for Orion with Mali G720 GPU. My board is equipped with 32 GB RAM and system is hosted on a 1TB Patriot P400 Lite NVMe drive. For monitoring, I simply rely on sbc-bench monitoring for temperatures and used two tasks to load the SoC, running a packer in parallel concurrently with glmark2 to somewhat stress the GPU at the same time, so I started these three commands in three shells at the same time

sbc-bench.sh -m
glmark2 --fullscreen
xz -9 -T 12 Fedora-Workstation-41-1.4.aarch64.raw

I also use a small watt-meter at the plug of the ATX PSU to check consumption

IMG_20250816_121004
IMG_20250816_121004.jpg4000×3000 3.52 MB

#22

Here is the original layout, with Radxa stock cooler. Sorry for the cable management, I kept things simple for tests

IMG_20250816_103808
IMG_20250816_103808.jpg1500×2000 1.11 MB
IMG_20250816_103839
IMG_20250816_103839.jpg3000×4000 3.72 MB

I have no idea of the thermal paste used by Radxa, but it was applied in a generous way (it was not a pad). I realized my snapshot was not focused correctly, but here are the unmounted parts (backplate screwed)

IMG_20250816_104419
IMG_20250816_104419.jpg3000×4000 3.72 MB

#23

Now the new passive heatsink is in place, with a generous ‘nut’ of Noctua NT-H1 thermal paste. I forgot to show the back of the heatsink, but the contact surface is half of the total area, with two ‘grooves’ on sides, so the contact is on a centered rectangle. Also, the contact surface is a bit rough, at least compared to the Radxa cooler.
I chosed the horizontal orientation for the fins in my test, assuming they are roughly aligned with the case air flow. I think the vertical orientation is also possible, for pure passive conditions for instance, but the heatsink really touches the plastic CPU fan plug whatever the choice!

IMG_20250816_105309
IMG_20250816_105309.jpg1500×2000 996 KB

Placement is a bit tricky as you have to carefully place and push the heatsink while fixing it on the provided backplate

Finally, here are the results of my tests:

Room temperature: 24°C
Idle power (no significant difference with CPU fan): 21W

Radxa stock cooler

  • Idle temperature: 32~33 °C
  • Peak temperature: 50 °C (after 5~10 minutes load)
  • Max power: 35 W

Passive 2U cooler

  • Idle temperature: 28~29 °C
  • Peak temperature: 42~43 °C (after 10~15 minutes load) - see next picture
  • Max power: 32 W

IMG_20250816_111949
IMG_20250816_111949.jpg2000×1500 1.52 MB

The difference between idle and load power looks strange, there is only a 10-13 W difference I can’t explain, maybe the measurement is very imprecise, but I was expecting more… Also, the max power consumption was 3W higher with stock cooler at load, I can’t explain this either, so please do not take these numbers for granted, it is simply not very reliable. Temperatures may not be precise either, these come from the chip sensors, but I guess the comparison between the two coolers still makes sense.
Even if the thermal paste applied on the stock cooler is not very efficient and lower temperatures might have been achieved with the Noctua paste, the passive heatsink shows a much better behavior and I had to relaunch the packer for an asymptotic value: just 10 degree difference while the stock cooler allows almost the double.

So, if I have to choose between the two coolers, the passive solution is of course better with a decent case air flow. No doubt that the stock cooler is sufficient for most users, but in server conditions, I would not be confident in a small CPU fan.

So, yes, I definitively recommend this aluminium passive heatsink, except maybe in pure passive conditions (additional tests would be required).

I hope this can help…

#24

Nice! So this shows that such a heat sink is a great solution for a regular case like yours. I even think that the rear fan could be stopped if you turned the heat sink 90°, as hot air would then be sucked by the power supply above it.

Also I’m not much surprised that the original small fan can draw 3W at full speed. I don’t remember if I measured mine but this sounds realistic.