As per the title, what are the maximum and minimum operational temperature ratings of the Pi before it stops reliably working? Could this also depend on the SD card in use?
From the RPi FAQ:
What is the usable temperature range?
The Raspberry Pi is built from commercial chips which are qualified to different temperature ranges; the LAN9512 is specified by the manufacturers being qualified from 0°C to 70°C, while the AP is qualified from -40°C to 85°C. You may well find that the board will work outside those temperatures, but we’re not qualifying the board itself to these extremes.
It'll go way down to < -70°C according to the article: Raspberry Pi proven to be stable when submerged in liquid nitrogen.
My experience with Raspberry Pi 3: The SoC will start to throttle down at approximately 80 degrees Celsius, and will, in my experience, never allow itself to be warmer than 85 degrees Celsius. This is of course the core temperature - the temperature outside the chip will have to be much lower to facilitate efficient heat exchange.
While you (probably, don't take my word for it) cannot destroy the SoC by leaving it uncooled, the performance will be severely impacted. (Same goes for the power supply, BTW). In our lab, we started noticing frame drops and significant degradation of video processing capability, only to find out that 1) it got too hot without the heatsink 2) the voltage dropped below 4.6V due to 5V supply wires that were too long.
In any kind of extreme scenario, it is most likely that your processing power will decrease first, and other problems will appear much later, if ever. This can lead huge waste of time when trying to hunt down software bugs ("why is my program suddenly running so slowly?!?"), only to discover that the wires are too thin, or the heat sink is too small, so beware!
Regarding the low boundary, you should check all the components. I recently booted Raspi3 at -12C cold and the camera did not work (first time in weeks, but other times the temperatures over night were not so low). After 15 minutes of waiting I rebooted it and it started working normally.
Also, I think that the networking/USB chip on the board itself is not rated below 0 C. If you need such extremes, I suggest waiting for Compute Module 3, which will have range -20 to 80C, simply by not providing the problematic chip at all :)
I see the OP's question has been answered authoritatively, but here are my 2 cents worth of experience:
With the basic clear plastic no-fan enclosure and heat sinks the ARM AP runs at about 50C (122F), and my Pi3 works fine. When I take off the top part of the plastic shell the temperature drops to 47-48.
So my conclusion is that the enclosure is not causing any measurable harm in this regard.
The command to return the CPU temperature in stdout is vcgencmd measure_temp
I see in comments that
uhoh mentions that if you want to use the temperature in a Python program, the command
os.popen('vgencmd measure_temp').read() will return the textual version of the temperature number.
Here is the way I use Popen to get the temperature into an integer variable:
from subprocess import Popen, PIPE . . . cmd = 'vgencmd measure_temp' p = Popen(cmd,stdout=PIPE, stderr=PIPE, shell=True) stdout, stderr = p.communicate() CPUtemp = int(stdout) . . etc
The above is taken from this code:
This post shows how to use
fping in a few different ways even though the results come in as
Also includes a cradle-to-grave example which makes use of the data and plots it live as it comes in. It shows Python and gnuplot. We don't see enough of these whole-system examples here.
The following is a little outside of the question, but a general use case that might give some ideas.
This can be adapted to any kinds of inputs, gpio sensors, internet datas.
How to graph the CPU temperature overtime?
Gnuplot can graph datas in the terminal, does not require any X server and use very little ressources.
It works smooth even on the slowest raspberry pi's model 1/zero.
sudo apt install gnuplot
Script example to build a
temperature script to store the data overtime.
#!/bin/sh echo $(date +%s ; cat /sys/class/thermal/thermal_zone0/temp) | tee >> temperature.plot
Give execution rights to this script:
chmod +x temperature
Detach and run in 1s loop till next reboot:
nohup watch ./temperature &
Later, graph the datas:
gnuplot -e "set terminal dumb $(tput cols) $(tput lines);plot 'temperature.plot' using 0:2 with lines"
This is a barebone example, temperature in Celsius * 1000, and seconds since the start, to be extended in your own scripts suite.
To kill the
Happy hacking ;)