Raspberry Pi 4 temperature

[martinrowan]

Might be some mileage in dropping the ARM idle frequency, in config.txt

arm_freq_min=300
or
arm_freq_min=150.

If anyone want to see if that makes much of a difference might be worth a go. Tiny performance hit as the CPU ramps up when started to load, but in my testing in the open air, seemed to get a 1deg drop in idle temps with 150.

Dropping the idle frequency to 150, showed a power draw reduction of ~40mA, to ~700mA. Resulting in very slightly lower temperatures, but nothing to get close to the ~480mA of an idle Pi3.

Disabling WiFi, Bluetooth and not using an ethernet connection either. Booting to a CLI not desktop, the current draw dropped to ~600mA.

[lfarrell12]

I expected this. I had a NanoPi practically melt on me 2 years ago so learned the lesson.
I couldn't get the Pimoroni fanshim so got the heatsink. Its about 65C with that.
I have a spare 30mm fan somewhere I need to crimp suitable pins for tomorrow. I am guessing that will help.

Generally most of the SBCs faster than the Pi3B+ either came with heatsink (and sometimes fan) or offered them as an option. Surprised the heatsink wasn't more strongly recommended as it will get hot.

[martinrowan]

Ok searching around I found some other sources:

• Raspi.tv Power measurements
• Core Electronics Pi3 vs Pi4 benchmarks
• Raspberrypi.org Benchmarks

All these sites, including raspberrypi.org, show the significant power usage increase between Pi 3 and Pi 4. Indicating this is normal and expected. So really it is that the new Pi 4 in an enclosure needs active cooling. So why on earth release the official case without it, when the Pi can't work properly in such a constrained environment.

[Imperf3kt]

I'll play devils advocate and suggest somebody didn't do their job properly. It's just that simple.

The case was probably designed before the Pi4b was fully tested, leading to what we see here.

Where the blame for that lies, is anybody's guess, but somewhere along production, somebody neglected the additional heat and the case designer didn't accommodate for it.

[jamesh]

Ok searching around I found some other sources:

• Raspi.tv Power measurements
• Core Electronics Pi3 vs Pi4 benchmarks
• Raspberrypi.org Benchmarks

All these sites, including raspberrypi.org, show the significant power usage increase between Pi 3 and Pi 4. Indicating this is normal and expected. So really it is that the new Pi 4 in an enclosure needs active cooling. So why on earth release the official case without it, when the Pi can't work properly in such a constrained environment.

The A72 use more power because they are twice as powerful. Simple as that. Not sure why people would think that would not be the case.

We actually have some ideas to get the temperature down, currently being tested. My open air Pi is running about 6degs cooler with some test firmware (not the USB firmware, already have that installed). Cannot say that is what will come out of the other end of testing, but looks promising.

As for the case design, not my field. Never used cases.

[jimmyken]

Ok searching around I found some other sources:

• Raspi.tv Power measurements
• Core Electronics Pi3 vs Pi4 benchmarks
• Raspberrypi.org Benchmarks

All these sites, including raspberrypi.org, show the significant power usage increase between Pi 3 and Pi 4. Indicating this is normal and expected. So really it is that the new Pi 4 in an enclosure needs active cooling. So why on earth release the official case without it, when the Pi can't work properly in such a constrained environment.

The A72 use more power because they are twice as powerful. Simple as that. Not sure why people would think that would not be the case.

We actually have some ideas to get the temperature down, currently being tested. My open air Pi is running about 6degs cooler with some test firmware (not the USB firmware, already have that installed). Cannot say that is what will come out of the other end of testing, but looks promising.

As for the case design, not my field. Never used cases.

It's a good idea from software. Thank you for your efforts.
If I can ask a question to you, stuffs, I humbly request you that:
"Why Pi 4 becomes ahead of producing schedule in 2020 without enough product testing by consumers?".
Heat-treating is one of items in product testing by consumers.
And this issue may cause your product liability.
I'm very concerned about this issue as a RasPi user and a possible donor.

[martinrowan]

The A72 use more power because they are twice as powerful. Simple as that. Not sure why people would think that would not be the case.

James, sorry I don't think it's as clear cut as you make out.

• A Pi 4 idling consumes more power than a Pi 3, even when running at a lower frequency and even when booted to a CLI not desktop. Here we're not pushing data through the new faster ethernet, nor the latest USB 3 bus. It's running the basics of the OS and allowing the CPU to run in a power-saving mode.
• Newer and Faster does not mean it needs to consume more power and generate more resultant heat. I'm glad the 165W Xeon processors of the past are long gone. You can see that power consumption is now remaining relatively flat https://www.karlrupp.net/2015/06/40-yea ... rend-data/ compared to processing capabilities.
• A move from 40nm to 28nm is normally associated with an increase in performance and decrease in power consumption.

We actually have some ideas to get the temperature down, currently being tested. My open air Pi is running about 6degs cooler with some test firmware (not the USB firmware, already have that installed). Cannot say that is what will come out of the other end of testing, but looks promising.

As for the case design, not my field. Never used cases.

You are producing a "Desktop Kit", marketing it as a "PC Replacement" (Eben), the addition of 2 HDMI ports, this version of the Pi is being pitched as a Desktop. A scenario that will see the Pi 4 in an enclosure, not left in free air. Using the official case, the metal shielding of the USB and Ethernet ports get really hot to the touch, if this isn't addressed I'm sure someone somewhere will manage to burn themselves on these external connections. (I've put the case back on mine and will try and measure the shield temperatures tonight)

I accept the case design may not be your area, but some team did define or measure the power and thermals of the board, sharing this information with the team designing the new case so that it would work. Then, of course, there would be all the testing of the final productised solution to ensure the requirements were met and the system worked together properly.

"A project like Raspberry Pi 4 is the work of many hundreds of people, and we always try to acknowledge some of those people here.

This time round, particular credit is due to James Adams, who designed the board itself (you’ll find his signature under the USB 3.0 socket); to Mike Buffham, who ran the commercial operation, working with suppliers, licensees, and resellers to bring our most complicated product yet to market; and to all those at Raspberry Pi and Broadcom who have worked tirelessly to make this product a reality over the last few years." - (Eben -Raspberry Pi 4 on sale now from $35)

It looks like these questions should be aimed squarely at JamesA and MikeB, but I doubt they'll be reviewing the forum or various blog posts. When we had issues with the PoE HAT, you pulled it from sale (eventually). I think the Pi 4 case and all the SKUs that include should be pulled from sale (once suppliers have access to more stock) until the products have been tested and actually work together. Or, is that the reason they are all awaiting stock, that the case has been pulled, but it's not being disclosed openly until there is a solution? If there is a problem, do what you did with the PoE Hat and publically admit it, even before you have a full solution.

I really hope some tweaks in the works can bring down the temperatures, but it's going to have to be by a significant amount.

[Jonovich]

Does the Pi4 memory size make any notable difference in the CPU temperatures recorded under various loads/circumstances? Anyone got any comparable stats?

Re the official Pi case, some well placed holes cut into the lid and sides I presume will help with cooling, whilst maintaining a good degree of protection? This being an enthusiast/hack device should lend its owners to attempt modifications like this?

The increased temperatures were well known about prior to launch, else the first set of launch mods would not have been cooling fans and heatsinks.

Jono

[lb]

We actually have some ideas to get the temperature down, currently being tested. My open air Pi is running about 6degs cooler with some test firmware (not the USB firmware, already have that installed). Cannot say that is what will come out of the other end of testing, but looks promising.

So that means the 3-5 °C with the VLI firmware plus around 6 °C on top? So around 10 °C total? Sounds promising. How much reduction is that in terms of current?

[gkreidl]

The problem with all those closed plastic boxes, official or not, isn't really new. Starting with the RPi 3B additional cooling was always needed if you didn't want to lose processing power by throttling.

My 3B uses a small heat sink and is in case with open sides, mounted sideways for better airflow. If I run a heavy load for a longer time (converting a video with HandBrake, for example) it may throttle back a bit, but never below 11000 MHz.

My 3B+ is running in a FLIRC case. With heavy load it may switch back to 1200 MHz. I've never observed further throttling.

My RPi 4 (4GB) will also run in a FLIRC case. I won't order one, until the FLIRC case becomes available.

[Montala]

My RPi 4 (4GB) will also run in a FLIRC case. I won't order one, until the FLIRC case becomes available.

It may not actually be availble yet, but is available for pre-order now at $11.20 from https://flirc.tv/more/raspberry-pi-4-case

A Pi 4 specific case is also available for pre-order from Argon ONE at the higher price of $25.00 on https://www.argon40.com/argon-one-pi4.html

I have no connection with either company, but am posting this in the hope that it will be of interest to those of you who are concerned about this problem. Other 'solutions' will, I am sure, also be available before long!

[gkreidl]

My RPi 4 (4GB) will also run in a FLIRC case. I won't order one, until the FLIRC case becomes available.

It may not actually be availble yet, but is available for pre-order now at $11.20 from https://flirc.tv/more/raspberry-pi-4-case

A Pi 4 specific case is also available for pre-order from Argon ONE at the higher price of $25.00 on https://www.argon40.com/argon-one-pi4.html

I have no connection with either company, but am posting this in the hope that it will be of interest to those of you who are concerned about this problem. Other 'solutions' will, I am sure, also be available before long!

I live in Germany and have a supplier who can deliver all parts, including the FLIRC case, as soon as it is available.

[PhilE]

Hi all,

Following a successful trial with our alpha testers, we are releasing a new firmware build for the USB3 host adapter that should save about 300mW and help the Pi4 to run cooler, which in turn improves performance in compute-intensive tasks.

Download the archive from here: https://drive.google.com/file/d/1PXwrnh ... sp=sharing
With the zip file on the Pi:

Code: Select all

$ unzip vl805_update_0137a8.zip
$ chmod a+x vl805
$ sudo ./vl805 -w vl805_fw_0137a8.bin
$ sudo reboot

The zip file includes the current shipping firmware (013701) in case you want to revert:

Code: Select all

$ sudo ./vl805 -w vl805_fw_013701.bin
$ sudo reboot

By the way, if you run "sudo ./vl805" with no parameters it displays the version of the currently running firmware. Note that this may differ from what's in the EEPROM if you've just reprogrammed it but not rebooted.

[Imperf3kt]

To be perfectly fair, the Pi4b was said to be capable of being used as a desktop PC. If we look at any desktop PC from the last two decades, every single one of them has a heatsink and fan, why should the Pi be different

[jamesh]

To be perfectly fair, the Pi4b was said to be capable of being used as a desktop PC. If we look at any desktop PC from the last two decades, every single one of them has a heatsink and fan, why should the Pi be different

TBH, given the power available, it's amazing that is can run without a heatsink/fan at all. But it does, it keeps itself within its parameters, doesn't blow up. To get best performance a small fan would certainly be my preferred option, but I don't have one on the open air one I have on my desk and it works fine.

[Gavinmc42]

Been running this 4B1 as a desktop for the last 4 days.
Has a 20 x 20 x 5mm heatsink, no fan.
Sits at 55C with 8-10 browser tabs open, gets upto 64C when watching YT's.
But it is not cased, I know my 3B+ run 10C hotter in the original Pi cases.
I expect the 4's will too.

My Celeron box runs from 45 to 95C and has fan, three fans in total.
A small heatsink and tiny fan on enclosed 4's will be my Desktop 4 solution.
Just a heatsink if not enclosed should be enough, but I need to wait till summer, it's winter here now.

[Aardappeltaart]

The A72 use more power because they are twice as powerful. Simple as that. Not sure why people would think that would not be the case.

Probably because they have phones that became more powerful AND more energy efficient, while not running any hotter in the last years.

I already dubbed the official case a `prefab sauna`, because the RPI is a little sweaty guy, running here 60° while idling on Raspbian Lite, nothing attached.

Putting the little giant in its home would raise his idle temperature another 10 degrees, and 70° is not a nice temp to start any job, you better relax, I would say.

That said, I did run the CNX-software 7zip (CBC-bench) (throttle) benchmark and my Pi in an ambient 25° room did not throttle much. It ran better without heatsink than the CNX test with heatsink.

I had my Pi lying flat on a wooden table, I decided to add some spacers, and idle temperature dropped to 53°.

With the new firmware and other thermal optimizations in the pipeline, I'm quite confident the RPI4 will be up to most jobs without heatsinks or fans.

It just deserves a bigger house to breath. (Or fancy some other places, like hanging it next to your drying socks.)

[jahboater]

TBH, given the power available, it's amazing that is can run without a heatsink/fan at all. But it does, it keeps itself within its parameters, doesn't blow up. To get best performance a small fan would certainly be my preferred option, but I don't have one on the open air one I have on my desk and it works fine.

Running the nasty "cpuburn" stress test, the old Pi3 would crash, the Pi3B+ would throttle, the Pi4 runs happily at full speed for long periods - even when over clocked.

That's in free air (mounted on edge).

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!

[jahboater]

I had my Pi lying flat on a wooden table, I decided to add some spacers, and idle temperature dropped to 53°.

Try mounting it vertically - it improves the convection cooling and you should see a small drop in the temp.

The new firmware should help too - see the post above.

[RaTTuS]

just as a test I did a
while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 10; done& stress -c 4 -t 900s
Which gave
https://pastebin.com/9zGtFR1e [it did thermal throttle after awhile but the temps started wiht
temp=52.0'C

having got 2x 2p's with a bit of thermal gloop between them and some insulating tape round the edge and some more gloop on the SOC , I placeds the 2x2P on top and
immediately got
while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 10; done& stress -c 4 -t 900s
[1] 630
stress: info: [631] dispatching hogs: 4 cpu, 0 io, 0 vm, 0 hdd
frequency(48)=600117184
temp=44.0'C
frequency(48)=1500398464
temp=51.0'C
...

though it did hover around 77/78 deg [didn't throttle in 4 mins ]

[lb]

Hi all,

Following a successful trial with our alpha testers, we are releasing a new firmware build for the USB3 host adapter that should save about 300mW and help the Pi4 to run cooler, which in turn improves performance in compute-intensive tasks.

Thanks. This definitely works and has a significant effect. In a half-open case, the idle temperature drops by about 3 °C for me.

[tvjon]

For the past couple of years or so, I`ve had 2 RPi3b`s mounted vertically as in attached pic`.

I recently replaced 1 of them with a Pi4 as I intend to do some big GCC builds with it.

The next Pi4 I buy though will run with no heatsink or fan, & I do not anticipate any problems.

$ vcgencmd measure_temp

temp=46.0'C

[Gavinmc42]

I don't have the official case yet, but few drilled holes and 25x25mm fan on the lid?
On the Pi3B+'s I pop the top off if they get a bit warm, cannot do that with the new Pi4 case

[jamesh]

VLI specific firmware debugging moved here https://www.raspberrypi.org/forums/view ... 8&t=244421

[Montala]

Hi,

It might be of interest to some users, but Pimoroni finally have stocks of a sizeable (40x30x5mm) heatsink in Pirate black, specifically for the Raspberry Pi 4, and which I am assured will fit into the "official" case. The cost of the heatsink is £2.40 (plus postage) which isn't too bad I suppose!

It might not make a lot of difference overall, but every little helps!