Pi4 Cooling

[jj_0]

73'C sounds pretty close to 80C when i understand throttling will start.
What happens after a few hours of watching hi-def video or browsing add and clickbait covered social media pages if its hitting 73c internally after only 1 hour or so??

Well, assuming idle temp of 41'C, 73'C is only at 82% of the usable (non-throttling range). I've let the test run overnight for ~10h20m. Highest temp is 71'C, highest temp of case is 49'C. Ambient temperature dropped to 24'C during the night but increased to 26'C during the night.

So it looks like at ~70'C the case is able to keep it stable.

ok you have convinced me... great response btw..

Thanks

I've done another run using cpuburn-a53. Not exactly the same circumstances, this time I used the perforated bottom of a Pi 3 flirc case and turned the case upside-down to let the warm air escape. The reason was that my Pi 4 smelled a bit like 'hot electronics'. The case only cools the CPU of course, not any other chips that might heat up.
Similar results as before with slightly higher temperatures that are reached quicker. I added CPU frequency as well to see if it's throttling. I leave it to the reader to spot when I opened a window which caused airflow and dropped ambient temperature from 26'C to 22'C


And finally, a run with overclock to 1.7GHz and ambient temperature a stable 30'C (in my utility closet). Now there is some throttling until I add a household fan in the mix.

[jj_0]

Ok, but how well does the built-in WiFi and Bluetooth work in a metal enclosure?

I haven't tested Bluetooth but Wifi seems to work fine from 1st floor to router on ground floor:

With case:

Code: Select all

pi@raspberrypi:~/Downloads $ iperf -c 10.0.0.254 -p 4711 -t 60 -i 10
------------------------------------------------------------
Client connecting to 10.0.0.254, TCP port 4711
TCP window size: 91.9 KByte (default)
------------------------------------------------------------
[  3] local 10.0.0.145 port 54912 connected with 10.0.0.254 port 4711
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  58.6 MBytes  49.2 Mbits/sec
[  3] 10.0-20.0 sec  66.0 MBytes  55.4 Mbits/sec
[  3] 20.0-30.0 sec  51.2 MBytes  43.0 Mbits/sec
[  3] 30.0-40.0 sec  61.5 MBytes  51.6 Mbits/sec
[  3] 40.0-50.0 sec  65.9 MBytes  55.3 Mbits/sec
[  3] 50.0-60.0 sec  65.5 MBytes  54.9 Mbits/sec
[  3]  0.0-60.2 sec   369 MBytes  51.4 Mbits/sec

Without case:

Code: Select all

pi@raspberrypi:~ $ iperf -c 10.0.0.254 -p 4711 -t 60 -i 10
------------------------------------------------------------
Client connecting to 10.0.0.254, TCP port 4711
TCP window size: 87.5 KByte (default)
------------------------------------------------------------
[  3] local 10.0.0.145 port 37026 connected with 10.0.0.254 port 4711
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec  50.6 MBytes  42.5 Mbits/sec
[  3] 10.0-20.0 sec  37.9 MBytes  31.8 Mbits/sec
[  3] 20.0-30.0 sec  54.9 MBytes  46.0 Mbits/sec
[  3] 30.0-40.0 sec  55.5 MBytes  46.6 Mbits/sec
[  3] 40.0-50.0 sec  54.8 MBytes  45.9 Mbits/sec
[  3] 50.0-60.0 sec  55.0 MBytes  46.1 Mbits/sec
[  3]  0.0-60.1 sec   309 MBytes  43.1 Mbits/sec

I'v run iperf multiple times both with and without case, results were better with case than without case. A bit unexpected, maybe the case shields the antenna from interference?

[raspipat]

I've got a Miuzei [*1] acrylic case with a little built-in fan.
It is a very nice little case, and the little fan works REALLY well.
Idle temps are 40C.
I've been running 30 minute tests using 'stress -c 4' and so far the fan easily keeps the peak temps at just under 60C
Ambient temps are 24C.

In order to keep the GPIO header pins free for use by other applications, I soldered a couple leads to the TP1 (+5) and TP12 (GND) pads on the underside of the PI and connected the fan to those leads.

[*1] - Miuzei case on Amazon;
https://www.amazon.com/dp/B07VDHTVGP

[rjbook51]

I simply took the top off the case and turned the pi upside down on a laptop cooler pad. Powered the cooler pad fan from a usb charger. Not elegant but I hide the pi behind my monitor anyhow. The only disadvantage is constant fan noise. It keeps my 3B+ under 46 deg C, full load, and so far my 4B under 65 deg C.

[CypherOz]

I had a heat sink - it only make a few degrees C improvement.

My Fan SHIM for Raspberry Pi arrived today.... idles low 30's and stress 100% x 4CPUs is mid 50's - yes!!

Cooling is a must - I'm new to RPi but not IT; makes sense, maybe RPi need to have a good cooling solution off the shelf

[DougieLawson]

Cooling is a must - I'm new to RPi but not IT; makes sense, maybe RPi need to have a good cooling solution off the shelf

They have that in the power over ethernet (POE) HAT.

None of my Raspberries have any need for a fan, so it's got to be an add-on.

[Fenlig]

I've been running a small Minecraft server with my Pi in a Flirc case, keeps the temps under 60c, but I can't touch the case for too long. I feel like I should get an active cooler

Would there be anything gained in removing the CPU lid so you can more directly mount a proper heatsink?

[Herbaldew]

My Flirc arrived a few hours ago.

Didn't run any kind of tests but real world results are (Running CumulusMX weather station software full time):

Pi 3B+ w/ Flirc 50
Pi 4B 2G w/ Flirc 56

*The 4B was running 68 without case

[MisterEd]

I just received a flirc case. Ran the same test as the guy from Explainingcomputers:

Ambient Temperature: 27.0'C
temp=41.0'C
temp=50.0'C
temp=52.0'C
temp=54.0'C
temp=54.0'C
temp=56.0'C
temp=57.0'C
temp=59.0'C

The hottest spot on the case that I could find was 41'C.

So this case not only looks beautiful, it also seems to work well.

I ran the same tests except I modified the script from Explaining Computers to output the results to a file. The following tests were run:
No case (17:26 ambient=25.2C)
No case w/ table fan (17:40 ambient=25.2C)
CanKit case w/o top (18:10 ambient=25.0C)
CanKit case w/ top (18:40 ambient=24.6C)
Flirc case (19:55 ambient=24.3C)
Note I have the 3 CanaKit heatsinks but decided not to install them because I was going to switch to a Flirc case.

[webbsmurfen]

And running it for 1h11m the highest temperature is 73'C. Highest temperature on the case is 47'C.

73'C sounds pretty close to 80C when i understand throttling will start.
What happens after a few hours of watching hi-def video or browsing add and clickbait covered social media pages if its hitting 73c internally after only 1 hour or so??

Well, assuming idle temp of 41'C, 73'C is only at 82% of the usable (non-throttling range). I've let the test run overnight for ~10h20m. Highest temp is 71'C, highest temp of case is 49'C. Ambient temperature dropped to 24'C during the night but increased to 26'C during the night.

So it looks like at ~70'C the case is able to keep it stable.

what program is used to create those graphs? i have used that script from explaining computers

[MisterEd]

73'C sounds pretty close to 80C when i understand throttling will start.
What happens after a few hours of watching hi-def video or browsing add and clickbait covered social media pages if its hitting 73c internally after only 1 hour or so??

Well, assuming idle temp of 41'C, 73'C is only at 82% of the usable (non-throttling range). I've let the test run overnight for ~10h20m. Highest temp is 71'C, highest temp of case is 49'C. Ambient temperature dropped to 24'C during the night but increased to 26'C during the night.

So it looks like at ~70'C the case is able to keep it stable.

what program is used to create those graphs? i have used that script from explaining computers

I know you are not asking me but I used the script from Explaining Computers also. I used Microsoft Excel to make my graphs because I more familiar with it. You should be able to do something similar with Libre Office Calc. As an exercise I might just try that also.

[tdimike]

Here I have graphed four runs, the first 3 are with a very low ambient temperature of 12.5 degrees, the final is in a much warmer room!

Stock clock and voltage
Unvervolted with Overvoltage= -8
1750MHz with overvoltage=1

[RonR]

Since we're posting graphs, here's one of a Pi4 with a software controlled small fan.

Looks like the software could stand a little improvement in its smoothing algorithm,

[Imperf3kt]

Looks like the software could stand a little improvement in its smoothing algorithm,

The fan is either on or off... not much smoothing possible. PWM modulation doesn't play well with brushless motors. The idea is to keep the temp below 70C with minimal fan operation.

Source?
10 seconds of googling and more than 2 decades of personal experience suggests otherwise
https://www.digikey.com.au/en/articles/ ... -dc-motors

[PeterO]

Source?
10 seconds of googling and more than 2 decades of personal experience suggests otherwise
https://www.digikey.com.au/en/articles/ ... -dc-motors

There is a big difference between the motor controller (called an ESC (Electronic Speed Controler) in modelling circles) performing PWM for speed control and something external performing PWM on the controlers power supply to attempt speed control.

One works, the other doesn't.

PeterO

[jj_0]

what program is used to create those graphs? i have used that script from explaining computers

LibreOffice Calc on the Pi 4 I find it quite usable.

[pfletch101]

I may eventually be able to tell you, but my Flirc case is currently even further back-ordered than my Pi 4.4!

Pimoroni has the Pi4 4GB in stock (actually they have all 3 flavors in stock at the time of this post).
https://shop.pimoroni.com/products/rasp ... 7087445075

Okdo has 4GB in stock as well.
https://www.okdo.com/us/shop/raspberry- ... oard-only/

Flirc told me they're shipping in 3-4 weeks, 2 weeks ago.

I should have taken your advice! Newark were showing my Pi as shipping on July 26. However, when I contacted them today to find out what was going on, the agent told me that they weren't expecting new stock until October!!!! Meanwhile, it looks as if everyone else is now out of stock, too. Grrrr!!

{EDIT} Canakit did apparently still have stock as of this morning, and are shipping me a 4 GB Pi 4, hopefully to arrive on August 2.

[Imperf3kt]

Source?
10 seconds of googling and more than 2 decades of personal experience suggests otherwise
https://www.digikey.com.au/en/articles/ ... -dc-motors

There is a big difference between the motor controller (called an ESC (Electronic Speed Controler) in modelling circles) performing PWM for speed control and something external performing PWM on the controlers power supply to attempt speed control.

One works, the other doesn't.

PeterO

Quite obviously if the fan includes speed control you should use that, but there is no reason why it won't work on a regular 2 wire fan.
The original claim was that

PWM modulation doesn't play well with brushless motors.

Nothing mentioned of fans that include their own speed control circuitry.

[exartemarte]

. . . There is however a different way to achieve variable 2-wire brushless motor control using the Pi's PWM pin.. More on this later.

Suddenly this thread is getting interesting - I'm looking forward to your next post.

(Not that I actually want to control the speed of my fan, but I'd like to know how.)

[Randomp]

Just got my Barch Designs Pi Holder machined aluminum case for my 4GB Pi 4. It's very pricey at $70 USD, but I figure well worth it for piece of mind on a machine running 24/7 and avoiding fan noise and dust.

Ran a sysbench stress test on it for 10 minutes and it did not go over 58C.



If you are interested in one, here is the link:
https://wickedaluminum.com/collections/ ... 5494435921

Makes the Pi into a tank, but I don't think it's for everyone. It is noted that it is likely to reduce Wifi and Bluetooth range, but I use neither of those, so I can't really confirm it in my use case.

I also ordered a Flirc case, but that has not shipped to me yet. I don't have too much hope that it can compete with this case, but it should do just fine for my 2GB Pi 4.

[jj_0]

Just got my Barch Designs Pi Holder machined aluminum case for my 4GB Pi 4. It's very pricey at $70 USD, but I figure well worth it for piece of mind on a machine running 24/7 and avoiding fan noise and dust.

Ran a sysbench stress test on it for 10 minutes and it did not go over 58C.



If you are interested in one, here is the link:
https://wickedaluminum.com/collections/ ... 5494435921

Makes the Pi into a tank, but I don't think it's for everyone. It is noted that it is likely to reduce Wifi and Bluetooth range, but I use neither of those, so I can't really confirm it in my use case.

I also ordered a Flirc case, but that has not shipped to me yet. I don't have too much hope that it can compete with this case, but it should do just fine for my 2GB Pi 4.

I can't decide on whether it's beautiful or monstrous. But compared to the Flirc case I have this one has heatsinks for other chips as well as for the CPU. So better heat dissipation I think. At a price. ..

[PeterO]

Too short mark pulses will not allow enough time for the motor to spin up, but duty cycles of 25-100% in mark-space mode work well to provide variable fan speed. Fan blade inertia takes care of smoothing the spin and pulses are not audibly perceptible, nor does the motor squeal.

Sounds more like "blipping" an engine
https://www.youtube.com/watch?time_cont ... 6PnKUEFX8g
PeterO

[webbsmurfen]

Just got my Barch Designs Pi Holder machined aluminum case for my 4GB Pi 4. It's very pricey at $70 USD, but I figure well worth it for piece of mind on a machine running 24/7 and avoiding fan noise and dust.

Ran a sysbench stress test on it for 10 minutes and it did not go over 58C.



If you are interested in one, here is the link:
https://wickedaluminum.com/collections/ ... 5494435921

Makes the Pi into a tank, but I don't think it's for everyone. It is noted that it is likely to reduce Wifi and Bluetooth range, but I use neither of those, so I can't really confirm it in my use case.

I also ordered a Flirc case, but that has not shipped to me yet. I don't have too much hope that it can compete with this case, but it should do just fine for my 2GB Pi 4.

That one looks awesome. Did you noticed any difference in range, when it comes to using Wifi/Bluetooth

[Ultrabit]

I thought I'd share my experience with colling my Pi4 B+. Granted I've not done stress tests etc but still I thought someone might find it of use. I'm using the metal case in the photo below.

Case lid ON, no heatsink at idle 69 degrees
Case lid OFF, no heatsink at idle 61 degrees

Adafruit 15x15x15mm heatsink case lid OFF at idle 61 degrees (ie. makes no difference at all). Heatsink noted to have got very hot.

28x28x18mm heatsink from Ebay/RS case lid OFF at idle 54 degrees
28x28x18mm heatsink from Ebay/RS case lid ON at idle 62 degrees (a reduction of 7 degrees in total).

There's 0.5mm between the top of the heatsink and the case lid so it fits in there the quite nicely. I probably could have got a wider one but didn't want to cover the RAM chip impairing it's heat dissipation.







link to the heatsink at RS, also found on Ebay
https://uk.rs-online.com/web/p/products ... lsrc=aw.ds

[Randomp]

...

I can't decide on whether it's beautiful or monstrous. But compared to the Flirc case I have this one has heatsinks for other chips as well as for the CPU. So better heat dissipation I think. At a price. ..

Yeah
It's machined out of a chunk of aluminum, so there's a lot of thermal mass compared to the Flirc. Size isn't much bigger than the Flirc though.

I'll do a stress test with my Flirc Kodi edition case when I get that to try and compare.

I only recommend this beast if you really need a fanless setup that will be crunching numbers 24/7. Otherwise it's overkill and pricey.

...

That one looks awesome. Did you noticed any difference in range, when it comes to using Wifi/Bluetooth

Sorry, mine is set up as a headless server using Ethernet. But either way it's right next to my router, so WiFi would probably work just fine for me.