Raspberry Pi 4 temperature

[f4bix]

Got my Pi4 today together with the Fanshim and never got above 60°C so far.

During a one minute stress test it reached a max of 58°C.

[jahboater]

This Pi4 heat sink case looks effective - but is that a Pi3 in the picture?

https://www.ebay.co.uk/itm/254277995339

[jahboater]

I'm not convinced the Pi4 needs a heat sink most of the time and my Pi4 doesn't seem to throttle, but they do no harm (the passive ones anyway) so I stuck one of these on for fun:

https://thepihut.com/products/al-heat-s ... 30-30-10mm

Its narrower than Pimoroni one mentioned above but heavier and it comes with proper specified 3M-8810 thermal tape. Make sure the fins are aligned vertically.

The tape is important, there have been cases in the past of cheap heat sinks for Pi's with insulating tape!!!!

[Gomoto]

The testing firmware is not working on my system (root on ssd was no longer getting to the desktop, hang after the 4 raspberries). SSD Sata to USB 3.0 does not work correctly with the testing firmware on my pi 4. I recovered by reinstalling raspian to my boot ssd. Using temporarily SATA to USB 2.0 (10mbs versus 200 mbs when sata usb 3.0 is working). Changing root back to ssd and changing back to the older firmware.

System is working fine again.

be careful not to harm usb 3.0 connectivity while saving power.
(additional info I did a rpi-update yesterday in advance to my root to ssd move, this messed perhaps with the testing of the forum firmware today)

Further info if it is helpful:
SATA USB 3.0 now up again (after manual back to the older firmware from the zip and a unrecommended rpi-update)
pi@raspberrypi:~ $ sudo hdparm -t /dev/sda3

/dev/sda3:
Timing buffered disk reads: 650 MB in 3.01 seconds = 216.11 MB/sec
pi@raspberrypi:~ $

[Gomoto]

I broke my small rpi heatsink from my rpi3 and put it on my new new rpi4 (hitting 79C during chromium video but no throtteling or temperature warnings. The small heatsink is quite helpful.

[burkskurk]

I’m also breaking 80C on idle with the box. Coincidentally, I also use the standard white/red box in headless-mode with VNC and force-hdmi. But there is close to no CPU-usage.

Running apt-get upgrade while having an X enabled and also browser open pushes temps towards 88C.

Removing the lid brings the temps down to 84.

[f4bix]

I'm not convinced the Pi4 needs a heat sink most of the time and my Pi4 doesn't seem to throttle, but they do no harm (the passive ones anyway) so I stuck one of these on for fun:

https://thepihut.com/products/al-heat-s ... 30-30-10mm

Its narrower than Pimoroni one mentioned above but heavier and it comes with proper specified 3M-8810 thermal tape. Make sure the fins are aligned vertically.

The tape is important, there have been cases in the past of cheap heat sinks for Pi's with insulating tape!!!!

I think it depends on the workload. For xorg use, I‘d definitely recommend a fan or heat sink to avoid throttling. For headless workloads a heat sink is more than adequate.

[bencat]

Hello would a Raspberry Pi Power over Ethernet (PoE) HAT v2.0 work on the Pi 4 ? If so would this help with the heat issue . Apologies if this is a repeated question as i did search but could not find a similar question.

[RaTTuS]

Hello would a Raspberry Pi Power over Ethernet (PoE) HAT v2.0 work on the Pi 4 ? If so would this help with the heat issue . Apologies if this is a repeated question as i did search but could not find a similar question.

yes

[Imperf3kt]

Hello would a Raspberry Pi Power over Ethernet (PoE) HAT v2.0 work on the Pi 4 ? If so would this help with the heat issue . Apologies if this is a repeated question as i did search but could not find a similar question.

It will work (was stated in the Pi4 launch blog) and additionally is actually recommended by at least one engineer.

[yslupdates]

I have the same issue, I only run pihole on a raspberry pi 4 for now and I average about 70C with the official case on.

[Gavinmc42]

Debug powerdownIf the Cortex-A72 processor runs in an environment where debug facilities are not required for any of itscores then you can reduce leakage power by turning off the power to the debug unit in the PCLKDBGdomain.To enable the debug unit in the PCLKDBG domain to be powered down, the implementation must placethe debug unit on a separately controlled power supply. In addition, you must clamp the outputs of thedebug unit to benign values while the debug unit is powered down.To power down the debug PCLKDBG power domain, apply the following sequence:1. Activate the debug output clamps.2. Remove power from the debug PCLKDBG domain.

Found this in the A72 manual, if the debug unit has a separate power supply, it could be turned off to save power.
Probably lots of tweaks like this can be done to reduce power, hence heat.
Don't have 4 full circuit schematic, but the basic one does not show this.
Maybe we need a 4B+ with debug power saving etc?

[Exaga]

The PI4 will run hotter - although its on a smaller process, it has much more powerful ARM cores, plus the HDMI output at 4kp60 adds quite a bit of heat. The ethernet is now on the main SoC also, that is another source of heat that wasn't there before. All in all, more performance requires more power requires more heat output.

We are still looking in to clock improvements to reduce heat load, its amazing how much reducing a clock where possible can reduce the overall temperature, it's a matter of finding settings that mean everything still works.

So, don't use hdmi_enable_4k=1 unless using 4kp60, that can make quite a difference.

James, I know the horse has already bolted on this one, and I know it's still relatively early days, but...

In view of what you're written, if RPi engineers are aware that the ARM A-series [Cortex-A72] CPU belongs to the "high-performance group" [and therefore uses quite a bit of power to operate and will generate quite a bit of heat in doing so], and after bench-testing and verifying that the device is suitable for release, and after your field-testers have put the RPi4 through it's paces and declared it suitable for release, and after Eben Upton has been quoted all over the news and media as remarking, "For the first time we provide a PC-like level of performance for most users.", and "No additional cooling will be required for 'light' use." (the latter seems to have vanished from the RPi website - not surprising as it's a little misleading), wasn't there any concerns that the device would not stand up to thermal scrutiny after it was released? I have to ask if the device was tested thoroughly by yourselves (or 3rd parties) prior to being released, and if so, why weren't any of the thermal issues highlighted during that process?

It's a bit of a PR disaster when the bold statements, and claims, and the inticing hype, that's been made about the Raspberry Pi 4 have turned out to be somewhat erroneous.

Very best of luck with your clock improvements. You will certainly struggle to tame the thermal dynamics of the beast you have created.

[jahboater]

and "No additional cooling will be required for 'light' use."

I believe that statement is correct. If you put it in a tiny un-ventilated case then that's the fault of the case design or poor choice of case - not the Pi4. (it is a bit sad that the official case seems to fall into that category). Who in their right mind puts a 10-15w device in a credit card sized case with no holes?

The idle temp is higher as noted above, but the heavy load temp seems to be more stable with less tendency to throttle than even the 3B+.

The Pi3, with a massive bolt-on heat sink, would throttle whilst running a stress test (but not in normal use).
My Pi4, in free air, with a small cheap stick-on passive heat sink, will not throttle running any stress test. Even cpuburn for an hour. As far as I am concerned its impossible to get it to throttle, which is way better than the Pi3. Probably all due to the 28nm shrink.

[jamesh]

The PI4 will run hotter - although its on a smaller process, it has much more powerful ARM cores, plus the HDMI output at 4kp60 adds quite a bit of heat. The ethernet is now on the main SoC also, that is another source of heat that wasn't there before. All in all, more performance requires more power requires more heat output.

We are still looking in to clock improvements to reduce heat load, its amazing how much reducing a clock where possible can reduce the overall temperature, it's a matter of finding settings that mean everything still works.

So, don't use hdmi_enable_4k=1 unless using 4kp60, that can make quite a difference.

James, I know the horse has already bolted on this one, and I know it's still relatively early days, but...

In view of what you're written, if RPi engineers are aware that the ARM A-series [Cortex-A72] CPU belongs to the "high-performance group" [and therefore uses quite a bit of power to operate and will generate quite a bit of heat in doing so], and after bench-testing and verifying that the device is suitable for release, and after your field-testers have put the RPi4 through it's paces and declared it suitable for release, and after Eben Upton has been quoted all over the news and media as remarking, "For the first time we provide a PC-like level of performance for most users.", and "No additional cooling will be required for 'light' use." (the latter seems to have vanished from the RPi website - not surprising as it's a little misleading), wasn't there any concerns that the device would not stand up to thermal scrutiny after it was released? I have to ask if the device was tested thoroughly by yourselves (or 3rd parties) prior to being released, and if so, why weren't any of the thermal issues highlighted during that process?

It's a bit of a PR disaster when the bold statements, and claims, and the inticing hype, that's been made about the Raspberry Pi 4 have turned out to be somewhat erroneous.

Very best of luck with your clock improvements. You will certainly struggle to tame the thermal dynamics of the beast you have created.

Your concern trolling has been noted. If the Pi4 does not fit your use case, don't use it.

[thatchunkylad198966]

I thought I'd register and join in.
I have a 4GB Pi 4 that's running quite hot. under load I'm getting: 80.0'C
idle is at: (as of right now) temp=60.0'C

I don't have the new USB3 firmware installed because of the problems that come with it. I also have a heatsink on my Pi 4.
It's hot; yes... but it's a better CPU and dual 4k. it's obviously going to be hotter than previous Pi's.

Anyway, everything is updated and working as expected _except_ the heat but I'm reading they'll be a firmware update to address that.

I do have a question though... when can we expect Chromium to have hardware acceleration?
That's the only bit for me (I don't ask for much) that's wrong.

Thanks for making me happy!

[phil995511]

Hi,

Personally I think that some of the problems of too high temperature come from the fact that the CPU and perhaps other components have been engraved in 0,28 microns whereas the prescriptions of ARM for this type of processor is 0,14 microns.

Another problem certainly comes from the small size of the case, the lack of ventilation slots to extract hot air and the material used, the plastic. The use of aluminum would certainly have made it possible to better evacuate excess heat.

The addition of passive radiators improves a little things but not enough.

I have +/- 63 ° C with the case open and +/- 70 ° with the case closed in IDLE mode (with frimware vl805_fw_0137a8)

As a personal I do not use Wifi or Bluetooth, I can disable them in the OS, but since there is no bios, I can not actually switch off. Would it be possible to boot Noobs on its SD card, to add a command line to physically turn off the Wifi / Bluetooth to hope less consume electricity and therefore less overheat ?

Best regards.

[Exaga]

Your concern trolling has been noted. If the Pi4 does not fit your use case, don't use it.

James,

When I first alerted people about the thermal issues with the Raspberry Pi 4 on 26 June 2019, I was labelled a "Troll" and a "Whinger" by some of them. Et tu Brute?

I regard this reaction and behaviour unbecoming of a professional of your standing. Please change your attitude. These are pertinent questions and were asked sincerely. I'm sorry to have bothered you for your insightful, esteemed, and valued opinion.

It's not my problem if the Rapberry Pi Foundation hasn't been forthcoming with the facts surrounding the RPi4 before it's release, or after. However, I have spent money on it. Therefore I have a legitimate right to comment and/or ask questions about the product I have purchased.

Good luck.

[EDIT] I've strapped a 28mm x 28mm x 20mm heatsink to the SoC and bolted a 30mmx7mm fan on top of it. Runs idle at ~35'C and maxes out all [4x] cores at ~45'C and will not rise above that. I'd say the thermal issue(s) on my Raspberry Pi 4 Model B are pretty much resolved. Thank you for your advice but I won't be taking you up on it. I'm questioning why an organisation as prestigous and prominent as the RPi Foundation could allow a product with very apparent thermal issues to be released to the public. I'm not saying the RPi4 is useless!

[jahboater]

As a personal I do not use Wifi or Bluetooth, I can disable them in the OS, but since there is no bios, I can not actually switch off. Would it be possible to boot Noobs on its SD card, to add a command line to physically turn off the Wifi / Bluetooth to hope less consume electricity and therefore less overheat ?

I use this in /boot/config.txt
Not sure if its still correct for the Pi4?

dtoverlay=pi3-disable-bt
dtoverlay=pi3-disable-wifi

If running headless, turn off HDMI by adding this to /etc/rc.local (saves a few ma).

/usr/bin/tvservice -o

Finally, add this to /etc/crontab to turn off the red power led.

@reboot root echo 0 > /sys/class/leds/led1/brightness

As for heat, mine idles at 44-45C and reaches 76-77C after an hour running cpuburn (a nasty stress test using NEON).
Small stick-on heat sink, no case.

[wrn]

I switched my old power hungry and noisy pc for a pi4 , and am doing my daily desktop activity with it : casual internet browsing, gimp editing , and a bit of compilation for stuff that I don't find distributed as packages.

After 3 days I can summarize things like that :

The pi4 sits on my desk table without a case, vertically, without a heat sink so far.
My room is about 25 °C

Casual internet over wifi like typing this right now with firefox ESR and listening to a radio stream, hanging around in matrix and doing some console stuff, the temp oscillates between 55 and 58 °C

Gimp editing ( mostly cropping, color balancing, some resizing , on ~3000 px images ) , makes it go between 55 and 68 °C

Compiling some libraries and some node stuff for a few hours, makes it climb over 70 ° , but it cools down back to 55 - 60 quickly as soon as compiling is finished.

My experience over the 3 days is very smooth for a daily light use, and temperature in my setup doesn't seem to be an issue.
I can say that I agree with ""No additional cooling will be required for 'light' use."

I would like to protect the raspberry and give it a housing, and I would like to keep enjoying the totally silent experience ( amazing feeling after 20+years of noisy fans in the room ) . I feel it would be a bad idea to put the little thing inside a plastic box without ventilation, so I guess I will try one of these heat sink/ enclosure combos when they are available, like the one mentioned by jahboater here : https://www.raspberrypi.org/forums/post ... #pr1491010

[jahboater]

so I guess I will try one of these heat sink/ enclosure combos when they are available, like the one mentioned by jahboater here : https://www.raspberrypi.org/forums/post ... #pr1491010

Please note my concern that it is described as a Pi4 case, but the picture appears to have a Pi3 in it!

The FLIRC case is very popular:
https://flirc.tv/more/raspberry-pi-4-case

Or the KODI version:
https://flirc.tv/more/kodi-edition-raspberry-pi-4-case

[wrn]

Please note my concern that it is described as a Pi4 case, but the picture appears to have a Pi3 in it!

Indeed the usb and ethernet are the other way around

[bensimmo]

I don't believe 99% of users need any additional cooling as long as the Pi4 has some ventilation.

The throttling limit has been raised from 60C on the 3B+ to 80C on the Pi4.

Wonderful - the best Pi model ever!

The problem with 80C on a metal surface is it can cause severe burns in a very short period of time. We're only talking a second or so.
The throttle should be lowered to 60C, even then it can burn but it takes longer and you'd have reacted by then.

At the moment cases are needed for user protection, which means it lacks ventilation and raises the temperatures, A catch 22 ?

[alphanumeric]

A Raspberry shaped / themed cutout in the lid would have helped, hint hint.
The first thing I did when I got the original official case was drill a grid of ventilation holes. I just taped some graph paper on top and drilled where the lines intersected. Then did the same on the bottom. If you take your time it doesn't look too bad when your done.

[bensimmo]

A Raspberry shaped / themed cutout in the lid would have helped, hint hint.
The first thing I did when I got the original official case was drill a grid of ventilation holes. I just taped some graph paper on top and drilled where the lines intersected. Then did the same on the bottom. If you take your time it doesn't look too bad when your done.

with one of them mechanical things that spins a cylinder with sharp bits, from ye olde times ?
or with the more modern method of setting up a laser cutter and enjoying the smell ?

P.S. I don't have a case, what plastic is it ?