Raspberry Pi 4 Heatsink vs Fan vs Flirc Case

[HawaiianPi]

I have used FLIRC cases for all my pi's. The new Pi4 FLIRC case has the same flaw as the older ones.

The flaw is in the cable. A 2x20 DuPont connector would fit, but most cable makers use the fat "IDE" drive style connector, and those will interfere with more than the Flirc. There is at least one pre-made cable that fits, but I don't have a link handy (try checking the Flirc support site, as I believe they suggest a cable)..

[PeterO]

I read elsewhere of a 1-4 dbm signal power reduction with some metal cases. Not too bad! 3dbm is equal to a 50% signal power reduction, which most receivers/transmitters can accommodate at short range.

0 dBm is 1 milliWatt.
3 dBm is 3dB more power than 0 dBm = 2 milliwatts.

What you mean is

I read elsewhere of a 1-4 dB signal power reduction with some metal cases. Not too bad! -3dB is equal to a 50% signal power reduction, which most receivers/transmitters can accommodate at short range.

PeterO

[jcyr]

I read elsewhere of a 1-4 dbm signal power reduction with some metal cases. Not too bad! 3dbm is equal to a 50% signal power reduction, which most receivers/transmitters can accommodate at short range.

0 dBm is 1 milliWatt.
3 dBm is 3dB more power than 0 dBm = 2 milliwatts.

What you mean is

I read elsewhere of a 1-4 dB signal power reduction with some metal cases. Not too bad! -3db is equal to a 50% signal power reduction, which most receivers/transmitters can accommodate at short range.

PeterO

That's true.

[JumpZero]

Hello,

I have the flirc case, a Pi4 4GB is inside.
I've done some stress tests recording temperature and frequency then ploting them in graphs.
Without and with overclocking: 1.5 then 1.75 then 2GHz
The test runs 15 minutes (900s) with the 4 cores 100% loaded.
Here under are 5 graphs
The 1st one is @1,5GHz without overclock. CPU loaded 100% 15minutes, temperature 65°C at the end of the test.
The 2nd is the same but at 14 minutes the load is stopped (killall stress) we can see temperature decreasing speed
The 3rd one is the same as 2nd (stress stopped at 14 minutes) but @1.75GHz
The 4th one is @2GHz without stop, but the test started with cold cpu at 40°C (shortly after power on)
The 5th one is @2GHz with a stop at 14 minutes but start with already hot cpu at 55°C on this test we can see after 10 minutes around 80°C the governor throttle the cpu: frequency goes back and forth between 1.33 and 2 Ghz keeping temperature stable around 80°C (alarm icon is blinking). At the end when stress is stopped the frequency goes to 666MHz and here too we can see the temperature decreasing speed.

5in1.png (35.43 KiB) Viewed 2238 times

And here is the code used in this test

Code: Select all

#!/bin/bash
# stress et gnuplot must be installed for this script to run
echo " Stress test - Press q to stop. Otherwise the test lasts 900s"
echo " Warning Ctrl-c doesn't stop the process stress. Use q"
echo " Printing and recording temperature and frequency every 5s"
echo " A graph is built at the end of the test"
echo " Press a key to start"
read -rsn1
stress -c 4 -t 900s &
CONTINUE=""
start_time="$(date -u +%s)"
origin_time=$start_time
total_time="$((0))"
rm -f TestResultClock.txt TestResultTemp.txt TestResultTime.txt
echo "total time: $total_time second" | tee TestResultTime.txt
vcgencmd measure_clock arm | tee TestResultClock.txt
vcgencmd measure_temp | tee TestResultTemp.txt
while [[ $CONTINUE != "q" ]] && [[ $CONTINUE != "Q" ]] && (( $total_time < 900 )); do
  read -t 0.25 -s -n 1 CONTINUE
  now_time="$(date -u +%s)"
  elapsed="$(($now_time-$start_time))"
  if (( $elapsed > 4 )); then
    start_time="$(date -u +%s)"
    total_time="$(($now_time-$origin_time))"
    echo "total time: $total_time seconds" | tee -a TestResultTime.txt
    vcgencmd measure_clock arm | tee -a TestResultClock.txt
    vcgencmd measure_temp | tee -a TestResultTemp.txt
    echo "Press q to stop. Press a to stop stress but continue recording"
  fi
  if [[ $CONTINUE == "a" ]] || [[ $CONTINUE == "A" ]]; then
   CONTINUE=""
   echo "Killing stress but test continues, temperature is cooling down"
   killall stress
   echo "stress killed"
  fi
done
echo "Test end"
killall -q stress
echo "stress killed"
echo "Done! Now let's build the graph."
cat TestResultTemp.txt | awk 'BEGIN { FS="=" } {print substr ($2, 0, 3)}' > TestResultTempTmp.txt
cat TestResultTime.txt | awk '{print $3}' | paste - TestResultTempTmp.txt > TestResultTempToGraph.txt
cat TestResultClock.txt | awk 'BEGIN { FS="=" } {print $2 / 100000000}' > TestResultClockTmp.txt
cat TestResultTime.txt | awk '{print $3}' | paste - TestResultClockTmp.txt > TestResultClockToGraph.txt
gnuplot -e "reset; set terminal png size 800,600; set pointsize 0.01; \
set output 'Temperature.png'; set title 'CPU load test';
set xlabel 'time (seconds)'; set ylabel 'Temperature(°C) / Frequency(MHz*100)'; plot \
'TestResultClockToGraph.txt' u 1:2 title 'Frequency(MHz*100)' with linespoints, \
'TestResultTempToGraph.txt' u 1:2 title 'Temperature(°C)' with linespoints"
rm -f TestResult*
echo "Et voila, the graph is in Temperature.png."
echo "END"

--
Jmp0

[riccetto80]

On my RPi4B-4G setup mentioned above,

changing the blue thermal pad to thin copper sheet improved cooling further
idling at 36C, watching YT videos at 40C, maxing at 45C during apt install/update/upgrades CPU at around 80%

interesting i was seeing on aliexpress what case to buy to have good passive and/or active cooling and i see your case.

i found this photo of the passive version:



and the issue there is cpu and ram are too distant to the metal frame, and is impossible to make contact beetwheen the usb3 controller and the case.

can you please tell me how your active version look? i guess under the 2 fans the case have metal, so the active version have metal much closer to the raspberry pi?

sorry for bad english.

[gregb49]

A most useful thread, thank you all, along with the mass of information at https://www.raspberrypi.org/forums/vie ... 8&t=243500. I'm most impressed with the Pi4, as a desktop replacement but am concerned over the heat issue. I know that the 'experts' say don't worry, but as an Electronic Engineer, I am uncomfortable running CPUs continuously at some of the temperatures recorded in these threads.

I have tried 3 RPi4 cases. Results of my non-scientific trial is as follows:

1. The official RPi4 case with a finned heat sink. It ran hot so I placed a small fan inside the case, using the existing ventilation holes. It now runs below 60degC, but is a bit noisy and the fan isn't secured. Works, but not recommended.

2. FLIRC. No fan, but a good thermal sink. My Pi ran up to 68deg and the case felt uncomfortably hot to the touch. I concluded that the case had insufficient thermal capacity (for my liking). I added ALDI hardware corner brackets ("ALDI Angle Set 4 piece" 2.99GBP) top and bottom, which although primitive and steel, brought the temp down into the high 40s, low 50s. As this is tucked away behind my desk, aesthetics are not important and goes to show how additional thermal capacity, albeit crude, can help.

3. Bruphny acrylic case. This included a fan and 3 heat sinks. I consider this the best of the three cases. Nicely made: The instructions are very clear: The fan can be run at 3volt and is almost silent. At the highest, it runs in the high 50s but normally in the 40s.

[LTolledo]

The picture of the armor case I've posted was from the amazon site....
below is how it looks now

RPi4B-4G_Armorcase.jpg (195.05 KiB) Viewed 2026 times

compared to the previous picture I've posted you'll notice that the fans are slightly raised
and the fan's power is not taken from the GPIO headers (this is my standard)
and opening it up.... notice I dont use the "insulator" over the SoC.

RPi4B-4G_Armorcase-open.jpg (200.56 KiB) Viewed 2026 times

with usual desktop use temp usually hover at 37C to 39C,
after watching 20minute 720p video via windowed VLC (seemed CPU intensive) temp rises and hovers at 44C to 45C, then drops to 37C to 39C in less than 15 secs after video end
fan noise measured at 22db measured 30mm above setup

[Thinkcat]

I just want to verify: Was there a difference in the height of the chips between versions of Pi 4? Or are all cooling solutions for Pi 4 for 1 GB, 2 GB and 4 GB versions equally?

[jcyr]

I just want to verify: Was there a difference in the height of the chips between versions of Pi 4? Or are all cooling solutions for Pi 4 for 1 GB, 2 GB and 4 GB versions equally?

I don't have a 2GB version, but the 1GB and 4GB appear the same.

[dom]

I just want to verify: Was there a difference in the height of the chips between versions of Pi 4? Or are all cooling solutions for Pi 4 for 1 GB, 2 GB and 4 GB versions equally?

No difference in height of sdram chip for 1GB/2GB/4GB.

[Konek]

Modded to fit heatsink for Pi4 4GB idle at 44C and when stress test 54C at room temp 26C with sysbench . If add a 30x30mm 5v fan will idle at 32C and at stress 37C.

https://drive.google.com/file/d/1-KxHRC ... p=drivesdk

[e-raser]

After overclocking the Pi 4 4GB for testing purposes to 1.75 GHz ARM frequency and running stable for a few days I set it to 2.0 GHz.

Stressberry benchmarks will follow in a few days, only notes so far:
- idle temperature with the DLIRC case is about 1 to 3 degrees Celsius higher than with stock clocking (1.5 GHz).
- to compensate the additional overclocking I attached an external USB powered fan and use UHUBCTL to control it dynamically according to the Pi‘s temperature, this works great so far.

[scaramonga]

3. Bruphny acrylic case. This included a fan and 3 heat sinks. I consider this the best of the three cases. Nicely made: The instructions are very clear: The fan can be run at 3volt and is almost silent. At the highest, it runs in the high 50s but normally in the 40s.

Just got this case, seems OK. Sysbench test on an hour loop run showed a MAX temp of 67c, clocked @ 2GHz, and with all the included (4) heatsinks fitted.

The real downside to this case, is the hdmi ports are too close to the case, resulting in the cable not fitting all the way into the socket, so dodgy connection, and a slight move of the unit, causes it to lose connection with the monitor.

[ctyldsley]

I'd love to see an all metal enclosure, like the Flirc case, with 3 heatsink pillars which is screwed fairly tight to the board. This would allow a thermal pad to be placed on each of the heatsink pillars for optimal heat transfer to the case itself. Then if the case had 2 low powered fans on the top running quiet.

Does such a case exist yet? I feel like this design would easily run at 100% in 2.0ghz for extended periods without any throttling.

[Imperf3kt]

I'd love to see an all metal enclosure, like the Flirc case, with 3 heatsink pillars which is screwed fairly tight to the board. This would allow a thermal pad to be placed on each of the heatsink pillars for optimal heat transfer to the case itself. Then if the case had 2 low powered fans on the top running quiet.

Does such a case exist yet? I feel like this design would easily run at 100% in 2.0ghz for extended periods without any throttling.

Reports suggest the Flirc case with no fans can already achieve this.
I cannot try for myself as I have neither Flirc case, nor desire to overclock my Pi4b as its already more powerful than my requirements.

[scaramonga]

Update on my complaint of the Bruphny case.

Took out 3rd layer from bottom and cut out a section for the HDMI slots, it's very thin plastic, so was easy to do. Now the HDMI cables fit perfectly The power 'C' socket can be left as is, as this is slightly higher than the HDMI ones, so the strength of the case is not compromised by leaving this section in place

Great cooling case, will now keep.

BTW...

Fan @ 5v, barely audible.

[ctyldsley]

I'd love to see an all metal enclosure, like the Flirc case, with 3 heatsink pillars which is screwed fairly tight to the board. This would allow a thermal pad to be placed on each of the heatsink pillars for optimal heat transfer to the case itself. Then if the case had 2 low powered fans on the top running quiet.

Does such a case exist yet? I feel like this design would easily run at 100% in 2.0ghz for extended periods without any throttling.

Reports suggest the Flirc case with no fans can already achieve this.
I cannot try for myself as I have neither Flirc case, nor desire to overclock my Pi4b as its already more powerful than my requirements.

I've not been able to test myself but seeing some benchmarks online I'd seen extended 100% cpu stress tests (when overclocked to 2.0ghz) see the Flirc case to throttle eventually which suggests the level of heat dissipation isn't enough to level out below 80c. Obviously 100% constant extended use isn't likely but it'd be nice to have a case which can comfortably manage that! I'm just not sure if one exists...

[splangie]

The Micro Connectors cpu fan is doing it for me, after I drilled a number of holes in the case. CPU is maxing at 68 running cpuburn-a53 test for 20 minutes. I haven't used and of the other tests.

https://www.microcenter.com/product/609 ... spberry-pi

[Brad Q]

My Flirc case typically idled at 50C in a 75F room. Under full extended(hour) load it never exceed 74C but ran up to 70C in a heartbeat. After a few months I have added a 12V 120mm fan running on 5V. One foot away it is dead quiet(the reason for using 120mm). It now idles at 36C in 75F room and never exceeds 44C. I am still thinking about adding some 10mm(ish) stand offs between the board and the case bottom. In theory that would add air flow and increase the surface area of the case heat sink. It would expose the inside of the case as well as the outside.