A short introduction
The project started two months before the final (not really, I’ll tell you why not below) photo was published.
Simply put, I wanted to do something cool with the Raspberry Pi 5. And definitely in Hardware. The first step was delidding CPU. And then I went into a drawer I hadn’t looked in 3 years and found spare parts from another cool project in there. The project by the way:
The project has been a great success in the past, too. Photos with this cluster were, and some of them are still on the tops of various Raspberry Pi publics.
And this is what I found in my stockpile, purchased years ago on Aliexpress
After precision fitting, I bent the tubes as follows
I also found these great copper disks I bought for dabbling with magnets. I adjusted them a little bit to the dimensions.
Securely fastened and filled with thermally conductive glue
I drew and printed out a simple mount, and all that’s left to do is put it together
And tighten the mount.
That’s it, ready to experiment.
To begin with, this heatsink is quite enough for Raspberry Pi 5 to work stably at medium load without additional active cooling and activation of the Peltier element.
The element Peltier is capable of consuming a very large amount of energy. Here, the lab PSU shows a short circuit (which is not really there, just not enough current) voltage 12V and current 1.3A, i.e. the cell consumption is about 15W. And it actually generates that much heat, and takes very little from the Raspberry Pi (I would say at that power, no more than 3-4W). I use a large horizontal fan (blower turbine) at the bottom to blow out the radiator. As you’ve already realized, it’s not about silence or saving energy.
I was ready for anything, but not for what awaited me next
I spent several hours testing and reading articles online, but in the end I was never able to overclock the Raspberry Pi 5 CPU over 3GHz.
I.e. 3GHz no problem. Everything works at speeds like in numerous tests on the internet. But when trying to put any (literally any) values above 3GHz, I get 3GHz or less. There has been no way around this.
The device worked completely stable.
Here are some funny findings:
Firstly, a great command to display CPU frequency, CPU temperature, voltage, trotting status, and RAM status, update once per second
watch -n 1 "vcgencmd measure_clock arm && echo && vcgencmd measure_temp && echo && vcgencmd measure_volts && echo && vcgencmd get_throttled
&& echo && free -h"
But for example, a value of 103100
This will cause the CPU frequency to drop to 0.125GHz
And from the performance test, it seems to be true
A value of 200 000 seems to be taking us back to the default
Quite interesting arm_freq=200100 Gives a frequency of 2.5 GHz.
So the arm_freq=200600 gives 3GHz
Tests also show a corresponding level of performance
But a value of arm_freq=200700 doesn’t make any difference.
For no reason at all, a value of 1 200 700 seems to be taking us back to the default again
So, there was no need to give the Peltier module more energy at all. I did not do it, as condensation can cause a lot of damage. And I have plans for this project, which I will share soon. See you on Twitter (X) and Instagram
I’m a systems engineer in JetBrains company. Uptime Lab founder. I’m glad to see you on my website! I hope you find my content useful. Please subscribe to my Instagram and Twitter. I post the newest updates there.