What is the maximum temperature of the CPU before it gets demaged ?

[LinuxTux23]

Hi folks,

I have a Raspberry Pi 4 with a fan, to avoid a overheating of the CPU.
The fan isnt so quietly. Thats why i wanna write a little python programm, that turn on the fan if the CPU gets to hot.

Now my question is, what is the maximum temperature wich the CPU can survive ?

Thank you,
LinuxTux23

PS: Sorry for my English I´m not a natíve speaker.

[timg236]

The CPU frequency is automatically reduced in order to avoid damage due to overheating so you don't need a fan to prevent damage. In the very worst case you will just see a performance reduction when running CPU intensive tasks e.g. compiling the Linux Kernel.

First, make sure you have the latest software which reduces power consumption. Install the latest Buster image from the downloads page or if you have already installed Raspbian

Code: Select all

sudo apt update
sudo apt upgrade
# Then reboot 

If you only want to run the fan when the CPU is getting warm then 80 C might be a reasonable threshold but lots you'll have to experiment with different workloads and will depend on heatsink/case or not etc

The bcmstat utility is very useful for determining if/when you want active cooling
https://github.com/MilhouseVH/bcmstat

[DougieLawson]

The true answer would appear to be the melting point of lead free solder and best of luck to the OP trying to run their Raspberry at the temperature.

The in-built throttling means you can't practically run over 85°C without using mythical overclocking settings that would be outside the spec for the processor.

[jbudd]

You don't actually need to write a program for this, it is possible to control a fan automatically using a device tree overlay and a transistor-

For example, this line in /boot/config.txt will set GPIO 12 high when the CPU temperature reaches 65C and low again when it cools down by 10 degrees.

Code: Select all

dtoverlay=gpio-fan,gpiopin=12,temp=65000

See https://raspmer.blogspot.com/2019/02/co ... o-fan.html for more details

[jahboater]

The in-built throttling means you can't practically run over 85°C without using mythical overclocking settings that would be outside the spec for the processor.

With the latest firmware, there seems to be no risk of even reaching 85C now.
My Pi4 has no fan, and the maximum temp is about 72C when running the cpuburn stress test for an hour.
Running huge compilations like building GCC brings it up to around 63C.
And that's overclocked!
Idles around 40C or less now.

Cooling problems and throttling are a thing of the past, unless you have a silly unventilated case.

Code: Select all

Time         ARM    Core    H264 Core Temp (Max)  IRQ/s      RX B/s      TX B/s
======== ======= ======= ======= =============== ====== =========== ===========
13:08:43 1600Mhz  500Mhz    0Mhz 39.92C (40.89C)  4,772       4,089      16,728
13:08:45 1600Mhz  500Mhz    0Mhz 39.43C (40.89C)     87         129         113
13:08:47 1600Mhz  500Mhz    0Mhz 39.43C (40.89C)     82          32         105
13:08:49  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     99         158         104
13:08:51  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     88         158         100
13:08:53  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     88          32         100
13:08:55  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     95         200         163
13:08:57  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     86         128         100
13:09:00  640Mhz  200Mhz    0Mhz 39.43C (40.89C)     89          32         100
13:09:02  640Mhz  200Mhz    0Mhz 38.95C (40.89C)     98         196         142
13:09:04  640Mhz  200Mhz    0Mhz 38.95C (40.89C)     92         158         100
13:09:06  640Mhz  200Mhz    0Mhz 38.95C (40.89C)     92          74         139
13:09:08  640Mhz  200Mhz    0Mhz 38.95C (40.89C)     86         128         100
13:09:10  640Mhz  200Mhz    0Mhz 37.97C (40.89C)     85          96         100

[jbudd]

With the latest firmware, there seems to be no risk of even reaching 85C now.

Is "firmware" different from the kernel version as shown by uname -r?
If it is, does it get updated by apt update? How do we tell what version we have?

My Pi4 does seem to run cooler than it first did. I'd put it down to the weather.

[DougieLawson]

There's a new piece rpi-eeprom that updates the bootcode stored in EEPROM on your RPi4B. The RPF/RPTL folks have now forced a package on every Raspbian Buster user to run that automagically at boot time (which will be all good fun until they ship bootcode with a catastrophic bug and then the brown stuff will make quick company with the air movement device).

[jamesh]

Just for completeness, the CPU's are tested to over 120degC during MTBF testing. They are still undamaged at that temperature, but we do not allow to get anywhere near that hot in the real world.

[jbudd]

Hmm I knew about changes for network booting, didn't realise the bootcode also related to thermal issues.

So "firmware" version is revealed by sudo vl805 (000137ab)
It is updated by apt update and possibly at every boot. (I don't like that but I guess it's important to keep bootcode and kernel in sync)

I disabled my fan and ran sysbench --test=cpu --num-threads=4 --cpu-max-prime=80000 run The maximum temperature reached was 74C and the CPU was not throttled.

Did I miss an announcement somewhere to say the temperature issue is now resolved? Good job RPT engineers anyway!

[HermannSW]

My Pi4 has no fan, and the maximum temp is about 72C when running the cpuburn stress test for an hour.
Running huge compilations like building GCC brings it up to around 63C.

Can you try it not overclocked?
I did compile OpenCV in >40min, and did log all relevant data, CPU frequency was 1.5GHz the whole time.
Maximal temperature was 81°C, average was 76.7°C:
https://www.raspberrypi.org/forums/view ... 1#p1561520

SInce I don't have fans, your and my measurements should match somehow.
Perhaps your overclocking reduced heat (should not), or you did not push the CPUs like OpenCV compilation did ...

[sparkyhall]

Board orientation makes a difference, vertical seems best for cooling and I recall jahboater saying his was vertically mounted.

[HermannSW]

Board orientation makes a difference, vertical seems best for cooling and I recall jahboater saying his was vertically mounted.

I need to try vertical next time, I thought air from above and below is good enough.
My Pi was 45°, only the edge of a small box did touch the bottom side.

You made a good point with vertical, I will hammer in Nx2 nails into book shelve 5.8cm apart.
That way I can just hang the PIs with GPIO up, power and HDMI connectors down, when running.
In addition to better cooling, they don't use space on working plate that way!
I did use that hanging concept before, but with wooden back which means no air from bottom (Wooden Board Pi):
https://www.raspberrypi.org/forums/view ... ?p=1550484

[Paul Hutch]

The in-built throttling means you can't practically run over 85°C without using mythical overclocking settings that would be outside the spec for the processor.

Not quite true. I've made multiple Pi3B+'s hit 94°C when running flat out in a 65°C oven, with clock throttling enabled. So when running in an environment that hot I include code to stop non-essential tasks after it reaches 85°C and to shut itself down at 90°C to try not to shorten the Pi's life too much.

[jahboater]

HermannSW,

Can you try it not overclocked?

Yes. It didn't seem to make any (easily) measurable difference.
But 100MHz is a small, safe, overclock for the Pi4.

Since I don't have fans, your and my measurements should match somehow.
Perhaps your overclocking reduced heat (should not), or you did not push the CPUs like OpenCV compilation did ...

Try this:-

Code: Select all

wget https://raw.githubusercontent.com/ssvb/cpuburn-arm/master/cpuburn-a53.S
gcc -o cpuburn-a53 cpuburn-a53.S
./cpuburn-a53

It uses NEON on all four cores. I don't know of anything else that raises the temp so quickly.
In the early days of the Pi3 it would crash in less than 10 seconds!

My Pi4 is mounted vertically in free air. There is a small stick-on heat sink on the SoC with the fins aligned vertically.
The unused sockets have short cables plugged in (because the PCB ground plane and the sockets dissipate heat).
With this latest software, I suspect the little heat sink is not really needed.

[HermannSW]

Try this:-

Code: Select all

wget https://raw.githubusercontent.com/ssvb/cpuburn-arm/master/cpuburn-a53.S
gcc -o cpuburn-a53 cpuburn-a53.S
./cpuburn-a53

It uses NEON on all four cores. I don't know of anything else that raises the temp so quickly.
In the early days of the Pi3 it would crash in less than 10 seconds!

I did -- thanks for that.

"top shows":

Code: Select all

top - 22:54:06 up 11 min,  3 users,  load average: 3.97, 2.51, 1.09
Tasks: 121 total,   5 running, 116 sleeping,   0 stopped,   0 zombie
%Cpu(s): 99.9 us,  0.1 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem :   3727.6 total,   3460.4 free,     97.8 used,    169.4 buff/cache
MiB Swap:    100.0 total,    100.0 free,      0.0 used.   3503.7 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND    
  907 pi        20   0    1724     64      0 R 100.0   0.0   4:54.18 cpuburn-a+ 
  908 pi        20   0    1724     64      0 R 100.0   0.0   4:54.18 cpuburn-a+ 
  909 pi        20   0    1724     64      0 R 100.0   0.0   4:53.80 cpuburn-a+ 
  906 pi        20   0    1724    312    256 R 100.0   0.0   4:53.97 cpuburn-a+ 
   40 root      20   0       0      0      0 I   0.3   0.0   0:00.90 kworker/3+
...

And only 80°C without fan ...

Code: Select all

pi@raspberrypi4B:~ $ vcgencmd measure_temp
temp=80.0'C
pi@raspberrypi4B:~ $ 

... after I did what I said -- running the Pi4 (left) that way works like a charme:



P.S:
Update, after 30 minutes of cpuburn still at 80°C!

Code: Select all

pi@raspberrypi4B:~ $ top -bn 1 -1 | head -15
top - 23:21:09 up 38 min,  3 users,  load average: 4.02, 4.01, 3.53
Tasks: 122 total,   5 running, 117 sleeping,   0 stopped,   0 zombie
%Cpu0  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu1  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu2  : 88.9 us, 11.1 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu3  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem :   3727.6 total,   3418.3 free,     99.8 used,    209.5 buff/cache
MiB Swap:    100.0 total,    100.0 free,      0.0 used.   3492.4 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
  906 pi        20   0    1724    312    256 R  94.1   0.0  31:56.13 cpuburn-a+
  907 pi        20   0    1724     64      0 R  94.1   0.0  31:56.36 cpuburn-a+
  908 pi        20   0    1724     64      0 R  94.1   0.0  31:57.14 cpuburn-a+
  909 pi        20   0    1724     64      0 R  88.2   0.0  31:55.69 cpuburn-a+
 1125 pi        20   0   10188   2756   2444 R   5.9   0.1   0:00.01 top
pi@raspberrypi4B:~ $ 
temp=80.0'C
pi@raspberrypi4B:~ $ 

[HermannSW]

After rearranging computer room I did hang Pi3B+ and Pi4B at bookshelf, again.
Free space on book shelf allows to place 7" 1024x600 DPI display above Pi4B
(connected with cable from RGB LCD HAT to connector behind display).
USB gooseneck camera mount allows to fixate (v2) camera
(raspivid preview displayed on top right border of HDMI0 display below, with "--dispnum 2" option).
Two small USB webcams connected to Pi4B as well are placed on nearby assortment box.
Unused sofar microHDMI2HDMI connector for HDMI1 can be used to connect old 42" plasma TV right of box:

[mahjongg]

Hi folks,

I have a Raspberry Pi 4 with a fan, to avoid a overheating of the CPU.
The fan isnt so quietly. Thats why i wanna write a little python programm, that turn on the fan if the CPU gets to hot.

Now my question is, what is the maximum temperature wich the CPU can survive ?

Thank you,
LinuxTux23

PS: Sorry for my English I´m not a natíve speaker.

The melting temperature of silicon is 1,414 °C

[W. H. Heydt]

The melting temperature of silicon is 1,414 °C

I rather suspect that the PCB will turn to ash rather before that temperature is reached.

[Imperf3kt]

The melting temperature of silicon is 1,414 °C

I rather suspect that the PCB will turn to ash rather before that temperature is reached.

I would expect nearby capacitors and other smd components to fail long before the PCB turns to ash.

[Heater]

At that chip temperature the bond wires will melt and everything stops.

If not the chip unsolders itself.

Caps and PCB could be quite unharmed.

Did I tell you about that time I plugged a 27032 EPROM into it's socket the wrong way around?

For I brief moment I could see the die inside glowing like the sun through the quartz window.

[cleverca22]

So "firmware" version is revealed by sudo vl805 (000137ab)

thats only the version of the usb3.0 controller firmware, which has zero control over how much power the main cpu is turning into heat

[X-Gen]

With the command below, you can see your CPU frequency:

Code: Select all

vcgencmd get_config arm_freq

At certain temperatures, your frequency will drop.
It is best to stay below these temperatures.
The maximum temperature that broadcom guarantees the CPU, is between 100C to 120C.
The maximum temperature the pi foundation capped the CPU to, is 85C.
Depending on what version of Pi you own, CPU overclocking will be lowered between 60 to 80C; which means you'll get lower performance out of it.
For that reason it's best to run below the temperature where the Pi lowers it's boost clock.

You can always connect the fan(s) to the 3,3V pin, instead of the 5V pin.
It would still do about ~50% of the cooling, at much lower RPMs and noise levels.