[Server-devel] do RPi 3 B+ still need heat sinks?

James Cameron quozl at laptop.org
Tue Apr 3 20:17:07 EDT 2018


On Tue, Apr 03, 2018 at 05:28:31PM -0400, Adam Holt wrote:
> On Tue, Apr 3, 2018 at 5:01 PM, James Cameron <[1]quozl at laptop.org> wrote:
> 
>     On Tue, Apr 03, 2018 at 10:33:12AM -0400, Adam Holt wrote:
>     > On Mon, Apr 2, 2018 at 12:36 PM, Adam Holt <[1][2]holt at laptop.org> wrote:
>     >
>     >     On Mon, Apr 2, 2018 at 10:44 AM, Adam Holt <[2][3]holt at laptop.org>
>     wrote:
>     >
>     >         1) Recap from 2016: heatsinks on the original RPi3 CPU are not a
>     >         cure-all when the RPi 3 is enclosed in a case that lacks
>     ventilation on
>     >         a hot day.  But CPU heatsinks Do Work when taking the plastic top
>     off
>     >         the of the original RPi 3.  The CPU throttling problem
>     "immediately"
>     >         goes away on such hot days...bringing the temperature back down
>     below
>     >         80C...as measured by the command:
>     >
>     >            vcgencmd measure_temp
>     >
>     >         2) It's snowing today right outside my window, so I can't easily
>     >         simulate a hot summer's day -- but can others who live in hot
>     >         environments report back their readings above, when running the
>     new
>     >         RPi3 B+ in various conditions?
>     >           ☆ With heatsink on CPU -- and without?
>     >           ☆ With motherboard fully enclose by a case -- and without?
>     >
>     >     My own results, with all 4 CPU's unloaded, in a chilly room:
>     >
>     >  RPi 3 with-heatsink-on-CPU / RPi 3 B+ / RPi 3 B+ with-heatsink-on-CPU
>     > 44-46C / 46-48C / 45-47C case's plastic top removed (w/o wind or active
>     > ventilation)
>     > 46-49C / 48-51C / 48-52C case's plastic top attached (contains 100+ small
>     holes
>     > on 1 end, allowing very little ventilation)
>     >
>     >     After I ran "yes > /dev/null &" 4 times, to fully load all 4 cores of
>     the
>     >     CPU:
>     >
>     >  RPi 3 with-heatsink-on-CPU / RPi 3 B+ / RPi 3 B+ with-heatsink-on-CPU
>     > 80-82C / 70-71C / 71-72C case's plastic top removed (w/o wind or active
>     > ventilation)
>     > 82-84C / 75-78C / 79-82C case's plastic top attached (contains 100+ small
>     holes
>     > on 1 end, allowing very little ventilation)
>     >
>     > RESULT: attaching a heatsink to the RPi 3 B+ CPU does not help.  It might
>     even
>     > make things a bit worse, hmm.
> 
>     Yes, your heatsink is no good.
> 
>     The B+ CPU has a heatsink or heat spreader already, that silver
>     coloured bevelled structure with the black dot and Broadcom logo.
> 
>     What is the shape, size, and attachment method for your added
>     heatsink?
> 
> It's the standard Canakit 7-fin aluminum heatsink shown here, attached to the
> CPU with its own basic 3M self-adhesive sticker:
> 
> [4]http://www.bestofjay.com/w/
> raspberry-pi-3-overclock-heat-test-flirc-case-vs-canakit-heatsink/
> [5]http://11986-presscdn-0-77.pagely.netdna-cdn.com/wp-content/uploads/2016/04/
> heat-sinks-installed.jpg

Given your results, probably wrong heatsink for the job.  No, I don't
know of a better one.

Theory; the adhesive won't make good contact with outer bevel, and
passive airflow above that outer bevel will be reduced.

Theory; bubble in adhesive if the heatsink was not angled slightly as
you pushed it down.

Theory; the thermal control system is very different to the previous
version, and this invalidates your test method.  The system uses
thermal mass and core frequency scaling, and you've changed the
thermal mass.  It's non-linear.

Report your kernel version; the corresponding Raspbian release has
changes for the B+.

See
https://www.raspberrypi.org/blog/raspberry-pi-3-model-bplus-sale-now-35/
for more technical detail on the thermals; note how the core frequency
varies and the thermal mass is a resource.

Suggest you use sysbench for at least 15 minutes before reading the
temperature.  The "yes" test is very weak.

> FWIW these same heatsinks do lower the temperature of the original RPi 3 by
> many degrees IF the CPU 's under medium load AND the top of the plastic case is
> removed :)

But irrelevant, as the CPU is a different model, has a flat surface,
and does not have a thermal control system that uses thermal mass.

Please remember; things change, and this breaks your knowledge.  You
must learn again.

> 
>     > In any case: this result is completely different the original RPi 3
>     (where past
>     > experiments have shown that a heatsink-on-CPU greatly lowers its
>     temperature,
>     > When The Case Is Open!)
>     >  
>     >
>     >     CONCLUSION: the RPi 3 B+ appears to be a real winner in CPU-intensive
>     >     conditions, even in a "wintry" conditions (room is about 15C, and it
>     >     continues to snow right outside the window).  The RPi 3 CPU is
>     supposed to
>     >     self-throttle at 80C, until it just about turns itself off at 85C. 
>     I'd
>     >     assume the RPi 3 B+ CPU does the same?  But do not know for sure. 
>     Thanks
>     >     to all who can add any similar data points, in warmer climate/
>     conditions.
>     >
>     >     CLARIFS:  Both RPi's were running a near-final prerelease of
>     >     Internet-in-a-Box 6.5 on Raspbian Lite to explore real-world
>     conditions.  
>     >     I waited 10+ min in all above 8 experiments before taking "steady
>     state"
>     >     temp readings.  Still, fluctuations in CPU activity (and temperature)
>     >     arise, even long after I ran "killall yes" to end the most intensive
>     CPU
>     >     activity.
>     >
>     >         3) Prelim thermal analysis of RPi 3 B+:
>     >
>     >         [3][6]https://cdn-images-1.medium.com/max/1600/1*0jU89Yu_
>     6miI-CODB
>     >         MuHAw.png
>     >         [4][7]https://medium.com/@ghalfacree/benchmarking-the-
>     raspberry-pi
>     >         -3-b-plus-44122cf3d806
>     >
>     >         --
>     >         [5]
>     >         [6]Unsung Heroes of OLPC, interviewed live @ [7][8]http://
>     unleashkids.org
>     >         !
>     >
>     >         --
>     >         Unsung Heroes of OLPC, interviewed live @ [8][9]http://
>     unleashkids.org !
>     >
>     >
>     > References:
>     >
>     > [1] mailto:[10]holt at laptop.org
>     > [2] mailto:[11]holt at laptop.org
>     > [3] [12]https://cdn-images-1.medium.com/max/1600/1*0jU89Yu_6miI-
>     CODBMuHAw.png
>     > [4] [13]https://medium.com/@ghalfacree/benchmarking-the-
>     raspberry-pi-3-b-plus-44122cf3d806
>     > [5] [14]https://medium.com/@ghalfacree/benchmarking-the-
>     raspberry-pi-3-b-plus-44122cf3d806
>     > [6] [15]https://medium.com/@ghalfacree/benchmarking-the-
>     raspberry-pi-3-b-plus-44122cf3d806
>     > [7] [16]http://unleashkids.org/
>     > [8] [17]http://unleashkids.org/
> 
>     > _______________________________________________
>     > Server-devel mailing list
>     > [18]Server-devel at lists.laptop.org
>     > [19]http://lists.laptop.org/listinfo/server-devel
> 
>     --
>     James Cameron
>     [20]http://quozl.netrek.org/
>     _______________________________________________
>     Server-devel mailing list
>     [21]Server-devel at lists.laptop.org
>     [22]http://lists.laptop.org/listinfo/server-devel
> 
> --
> Unsung Heroes of OLPC, interviewed live @ [23]http://unleashkids.org !
> 
> References:
> 
> [1] mailto:quozl at laptop.org
> [2] mailto:holt at laptop.org
> [3] mailto:holt at laptop.org
> [4] http://www.bestofjay.com/w/raspberry-pi-3-overclock-heat-test-flirc-case-vs-canakit-heatsink/
> [5] http://11986-presscdn-0-77.pagely.netdna-cdn.com/wp-content/uploads/2016/04/heat-sinks-installed.jpg
> [6] https://cdn-images-1.medium.com/max/1600/1*0jU89Yu_6miI-CODB
> [7] https://medium.com/@ghalfacree/benchmarking-the-raspberry-pi
> [8] http://unleashkids.org/
> [9] http://unleashkids.org/
> [10] mailto:holt at laptop.org
> [11] mailto:holt at laptop.org
> [12] https://cdn-images-1.medium.com/max/1600/1*0jU89Yu_6miI-CODBMuHAw.png
> [13] https://medium.com/@ghalfacree/benchmarking-the-raspberry-pi-3-b-plus-44122cf3d806
> [14] https://medium.com/@ghalfacree/benchmarking-the-raspberry-pi-3-b-plus-44122cf3d806
> [15] https://medium.com/@ghalfacree/benchmarking-the-raspberry-pi-3-b-plus-44122cf3d806
> [16] http://unleashkids.org/
> [17] http://unleashkids.org/
> [18] mailto:Server-devel at lists.laptop.org
> [19] http://lists.laptop.org/listinfo/server-devel
> [20] http://quozl.netrek.org/
> [21] mailto:Server-devel at lists.laptop.org
> [22] http://lists.laptop.org/listinfo/server-devel
> [23] http://unleashkids.org/

> _______________________________________________
> Server-devel mailing list
> Server-devel at lists.laptop.org
> http://lists.laptop.org/listinfo/server-devel


-- 
James Cameron
http://quozl.netrek.org/


More information about the Server-devel mailing list