With the “stock” Khadas 3705 Cooling Fan and New VIM Heatsink, the equilibrium temperature is around 51C, at 25C ambient, on the VIM3. This is again running the SoC at maximum load using the Antutu 8 stress test. Notice that due to the New VIM Heatsink’s small thermal mass, it reached equilibrium a lot faster than @RRZC777’s large passive heatsink.
This test only took 30 minutes, and the 3705 cooling fan was set to “automatic speed mode” from within Android. The sharp dips at the start of the test was myself placing my cold fingers onto the heatsink, notice that since my body temperature is around 37C, my fingers sucked heat out. The second sharp dip is myself restarting the test around the 15 minute mark.
30-minute Antutu 8 stress test on the Edge-V (RK3399), using @RRZC777’s large passive heatsink. I stopped the test at 30 minutes, as the SoC was experiencing significant thermal throttling, and since our aim is to avoid that, I saw no need to proceed further.
Ambient temperature was at 30C, and the likely equilibrium temperature would be 75-80C. This is already too hot for most people’s standards.
You’re welcome! Now my issue is: should I rerun the tests using CPU Throttling Test…or move on to Ubuntu? Ah choices choices. Temperature is higher today too, 28C (with the air conditioner, 30C is a little too high for my liking, my brain is starting to throttle)
Just to properly end-off this round of testing with Antutu 8, here’s the final result for Edge-V with active cooling (3705 fan + new vim heatsink), and using the 24W adapter. The fan was again running at “automatic speed mode”, and I note that it definitely sounds louder as compared to when it was stuck onto the VIM3.
This is what I made with the 35x35mm generic heatsink with a thermal pad under. The idea for a totally passive top heatsink would be this, but as a whole block of 70x35mm, 15mm tall instead of the 5mm height that the actual heatsink has (my addition is 10mm height). The problem I see with the bottom heatsink is not that it’s in bottom per se (although its a problem since hot air goes top) but the thickness of the thermal pad to avoid touching bottom connections with the metal plate. Metal has much more thermal conductivity than thermal pads, and the thicker it is the more it insulates, it should be only a layer to fill the gaps between metals alas thermal paste.
Or just the metal plate without the left heatsink part so we can add the heatsink we want that fits our user case?
I will tell you once I manage to fix the SPI flash booting (from serial debugging) since now my board it’s a paperweight not being able to boot from SD/EMMC. Waiting on another thread in Technical Support for some help
@karibu - im building a cluster - using small heatsinks on each vim3 with a fan over all of them. Will be building this weekend and will post pictures / design info
@Vladimir.v.v results of the testing with “CPU Throttle Test v1.3.4”. I’ll need your help interpreting the results, because I’m unfamiliar with this program.
The only thing I know is that the blue lines represent throttling - I think. I also recorded the SoC’s surface temperature with the probe, just to be sure.
VIM3L, 24W Adapter, @RRZC777’s Large Radiator (Underside), 45-Minutes, CPU Throttle Test
Overall, I feel the Antutu 8 Stress Test is more “stressful” than the CPU Throttle Test, especially on the Edge-V. You can see that during Antutu’s test the temperatures were a lot higher. Even though the ambient temperature was about the same: 29 - 30C (assuming heatsink-ambient temperature relationship is linear)
