The hardware I chose to use for this test is as follows:-
- Raspbery Pi 4 4GB
- ZkeeShop Aluminum Case Alloy Armor with Cooling Heatsink - found on Amazon https://www.amazon.co.uk/gp/product/B07VDYF41Q
- Noctua NF-A4x10 40mmx10mm Fan 5V
- Noctua NF-A4x20 40mmx20mm Fan 5V
- Alphacool Eisschicht Light thermal pad - 7W/mK - 1mm
- Cool CPU to Idle Fan Temperatures for 20 mins
- Start Stress test run for 30 mins
- create log of temperatures every 6 seconds over 1 min at 20 min mark
- Cool CPU to Idle
- Swap hardware if needed
- Cool CPU to Fan Idle Temps
- Dissemble Heat sink case
- Remove Thermal Pad
- Clean CPU IHS and hea tsink with Isopropyl aclohol
- apply new Thermal Pad
- Assemble Heat sink case
- Power on and start over
- Out the box no heat sink with and without a fan - Idle, Stress test and cpuburn-a53 test
- Heat sink with packaged thermal tape with and without a fan - Idle, Stress test and cpuburn-a53 test
- Heat sink with Alphacool 7W/mK thermal tape with and without a fan - Idle, Stress test and cpuburn-a53 test
All results can be found here https://1drv.ms/x/s!AnduNu9ZFdCnpzMHNlOgjrcV1n3e
Out the box with no heat sink
I started testing without the heat sink case as the heat sink case only come with 2 small squares of thermal pads. I started with Idle temps first straight out the box then used the 10mm thick fan followed by 20mm thick fan.
Code: Select all
Out the box - Idle Avg: 56.124°C, Stress Test: 77.463°C, cpuburn-a53: Failed due to thermal throttling Noctua NF-A4x10 - Idle Avg: 44.436°C, Stress Test: 59.843°C, cpuburn-a53: 64.713°C Noctua NF-A4x20 - Idle Avg: 42.399°C, Stress Test: 57.363°C, cpuburn-a53: 62.056°C
Heat sink case with packaged thermal pads
This was an interesting tests due to accidental findings. While recording the temp data to a spreadsheet i put the fan on it's side and it was blowing against the side of the Raspberry Pi on the GPIO side then i noticed the idle temperature drop below what i recorded so I decided to run an additional test for each senario, that being with the fan on top of the heat sink case and the fan to the side (GPIO side). I shose the GPIO side as it gave the best result.
Code: Select all
Heatsink stock thermal pad - Idle Avg: 45.808°C, Stress Test: 61.481°C, cpuburn-a53: 69.716°C Heatsink stock thermal pad + NF-A4x10 Top - Idle Avg: 43.418°C, Stress Test: 56.212°C, cpuburn-a53: 61.702°C Heatsink stock thermal pad + NF-A4x10 Side - Idle Avg: 42.266°C, Stress Test: 54.087°C, cpuburn-a53: 57.319°C Heatsink stock thermal pad + NF-A4x20 Top - Idle Avg: 41.957°C, Stress Test: 54.663°C, cpuburn-a53: 59.843°C Heatsink stock thermal pad + NF-A4x20 Side - Idle Avg: 40.451°C, Stress Test: 52.051°C, cpuburn-a53: 54.530°C
Heat sink case with Alphacool Eisschicht Light thermal pad 1mm - 7W/mK
Things got even more interesting in this test. When i switched the Raspberry Pi on and started my temp monitor script i instantly noticed the temps idling lower than the packaged thermal pads by a significant margin
Code: Select all
Heatsink Alphacool 7W/mK thermal pad - Idle Avg: 40.628°C, Stress Test: 48.420°C, cpuburn-a53: 54.707°C Heatsink Alphacool 7W/mK thermal pad + NF-A4x10 Top - Idle Avg: 37.618°C, Stress Test: 43.860°C, cpuburn-a53: 45.631°C Heatsink Alphacool 7W/mK thermal pad + NF-A4x10 Side - Idle Avg: 37.795°C, Stress Test: 43.727°C, cpuburn-a53: 46.295°C Heatsink Alphacool 7W/mK thermal pad + NF-A4x20 Top - Idle Avg: 36.777°C, Stress Test: 42.222°C, cpuburn-a53: 44.746°C Heatsink Alphacool 7W/mK thermal pad + NF-A4x20 Side - Idle Avg: 36.688°C, Stress Test: 41.691°C, cpuburn-a53: 43.772°C
The Alphacool Eisschicht Light thermal pad - 7W/mK, out performed the packaged thermal pads by a significant margin of between 9.3% to 26.05% (calculated on the averages) with the largest gains found on the more stress put on the CPU. A single 120mm x 10mm x 1mm Alphacool 7W/mK thermal pad strip only cos £6.99 and with this much of a margin compared to the thermal pads that come in the box with the case, its worth considering switching to a better thermally conductive pad especially when you want to keep the CPU cool.
I had an idea that the thermal pad that was packaged with the case was going to be a poor performer and the results are eye opening
The Noctua NF-A4x20 pefromed the best in every test involving fans and its really quiet at 5V but you can quieten it more by using 3.3V which i have not done any testing at 3.3V but with a quick check, at 3.3V it idles at 37°C-38°C and is extremely quiet.
When it comes to thermal pads, the provided thermal pads for the ZkeeShop case does not perform well to the point where cooling the case from the side was better than cooling it from the top. I believe this happens as there is direct airflow to the part of the case that contacts the thermal pad. With switching to a better performing thermal pad, cooling between the top or the side of the case showed negligible between 0.24% and 2.1% (less than 1°C). The data demonstrated that the packaged thermal pads has a lot lower thermal conductivity than that of the Alphacool Eisschicht Light thermal pad - 7W/mK
All data from my tests can be found here https://1drv.ms/x/s!AnduNu9ZFdCnpzMHNlOgjrcV1n3e
Testing to Continue
For now my testing has come to a halt but not finished as i do want to test 2 other thermal pads from Alphacool, 11W/mK and 14W/mK thermal pad and once i am done with thermal pads i do want to try some thermal paste as well. Thermal paste testing will require some modification of the heat sink case in order to reduce the gap between the case and CPU so that the case contact area is just touching the CPU.
The last ting i would like to do after testing between thermal pads and thermal pastes is to design my own custom heat sink case for the Raspberry Pi. The ZkeeShop heat sink case does perform well, its not a bad case but i have my own ideas when it comes to cooling the Raspberry Pi and would like to see if i could find a better design to improve cooling performance.
Sorry for the post being rather long, i just thought this was rather interesting, If you have any ideas on how to improve the testing, i am willing to hear it. I am thinking of doing all the tests again and log the temps every second for a more detailed test from idle to stress test for a set time back to idle again and maybe with overclocking as well to see how far it can be pushed with different thermal pads before throttling, but i can only do the tests 1 more time as i only have 1 thermal pad left that was included with the case. if you have any ideas, i would love to hear them.
Thanks for your time