BBAI runs extremely hot

1,322 views
Skip to first unread message

Paul Richards

unread,
Sep 25, 2019, 7:51:41 PM9/25/19
to BeagleBoard
I got 2 BeagleBone AI's from Mouser and started working with them today.

Both of my boards boot and run with the 08-03 build.

I haven't done anything to modify the contents of the filesystem, I merely booted and connected via Cloud9.

On both boards, the main CPU heat sink and the Kingston RAM module immediately
heat up and run incredibly hot.  I measured >80C on the CPU and >90C on the RAM module with my IR thermometer.

System load looked fine, nothing was running away and stealing all the CPU.

We intend to use these in lieu of BeagleBone Black modules for a project in the Keck telescopes and if they normally run this hot, it will never work.

S/N range is 4000167 1301 003491 1933


Is my experience atypical?  What temperature should these boards nominally run at?



Paul

Robert Forsyth

unread,
Sep 25, 2019, 9:52:43 PM9/25/19
to BeagleBoard
I upgraded the software as recommended by start guide and syslog kept reporting kernel: overtemp, shutdown.

Reflashing appeared to fix that, but mine is still slightly too hot to touch heatsink, heat sink about 50 C

The (Linux) CPU activity LED seems on a lot, when doing nothing (useful).

Dennis Lee Bieber

unread,
Sep 25, 2019, 11:10:44 PM9/25/19
to Beagleboard
On Wed, 25 Sep 2019 18:52:43 -0700 (PDT), in
gmane.comp.hardware.beagleboard.user Robert Forsyth
<robert.styles.forsyth-...@public.gmane.org> wrote:

>I upgraded the software as recommended by start guide and syslog kept
>reporting kernel: overtemp, shutdown.
>
>Reflashing appeared to fix that, but mine is still slightly too hot to
>touch heatsink, heat sink about 50 C
>
I'd done a kernel update -- on a booted SD card image -- and on reboot
it did the same... Scrolling temp warning and notice of shutdown (but
apparently spending so much time handling the temp warning it wouldn't
actually shutdown; had to pull the power).

-=-=-=-
Last login: Mon Sep 23 15:12:20 2019 from 192.168.1.66

Message from syslogd@beaglebone at Sep 23 15:32:10 ...
kernel:[ 26.752867] thermal thermal_zone1: critical temperature reached
(88 C), shutting down

Message from syslogd@beaglebone at Sep 23 15:32:10 ...
kernel:[ 26.776858] thermal thermal_zone0: critical temperature reached
(90 C), shutting down

Message from syslogd@beaglebone at Sep 23 15:32:10 ...
kernel:[ 26.880858] thermal thermal_zone2: critical temperature reached
(88 C), shutting down

Message from syslogd@beaglebone at Sep 23 15:32:10 ...
kernel:[ 26.976768] thermal thermal_zone4: critical temperature reached
(89 C), shutting down

Message from syslogd@beaglebone at Sep 23 15:32:10 ...
kernel:[ 27.040931] thermal thermal_zone0: critical temperature reached
(90 C), shutting down

<SNIP over 1000 such entries>
-=-=-=-

Had to reimage the SD card, but wasn't getting warnings from that
"stock" image. Didn't measure temps (I do have an old Ryobi IR
thermometer). I did notice the card edges were rather heated when I shut it
down and picked it up.

apt-get update/upgrade alone didn't cause the situation, seems to be
something in the kernel update -- a change in thresholds? Or some
calibration flaw?

I do think I should locate some fan that can mount on it (and here I
though an R-Pi 3B in a case was running warm -- my Pi-Star node typically
registers 50-53degC, and that is "amber" on the status page.

--
Dennis L Bieber

jks

unread,
Sep 26, 2019, 6:07:00 AM9/26/19
to BeagleBoard
Connecting with ssh I found that out-of-the-box if I ran top or htop the lxqt-panel process was taking a consistent 5% of one of the cores. After updating it increased to 9% (lol). A "killall lxqt-panel" put a stop to that and idle time returned to 99%. The GUI stuff shouldn't even be running without a valid HDMI HDID, right? Can't wait to get a console image working and leave all that GUI junk behind..

antiw...@protonmail.com

unread,
Sep 26, 2019, 5:08:29 PM9/26/19
to BeagleBoard
Looks similar to me. Updated to latest images and get:

 kernel:[  709.540231] thermal thermal_zone0: critical temperature reached (80 C), shutting down


Myrick Monroe

unread,
Sep 26, 2019, 5:08:30 PM9/26/19
to BeagleBoard
I had the same thermal problem and my board shut down by itself. I was able to solve this problem by placing an 80mm fan on top of the board. This fan has the usual 3 wires and I connected the red wire to 5V Sys (P9-7) and the black to GND in P9-1. The entire board now runs cool to the touch.  

Mike Brandon

unread,
Sep 28, 2019, 8:19:09 AM9/28/19
to BeagleBoard
@Paul - Would you be able to take a close up picture of the backside  solder points of the serial debug connector (3 pin next to USB C)? 

My AI was overheating too until it stopped booting all together. I was looking to monitor serial as I applied power and that is when I noticed what appeared to be an unclean solder point...its not shorted there is a 99kOhm resistance between the two points.


Mike

Dennis Lee Bieber

unread,
Sep 28, 2019, 11:56:13 AM9/28/19
to Beagleboard
On Sat, 28 Sep 2019 05:19:09 -0700 (PDT), in
gmane.comp.hardware.beagleboard.user Mike Brandon
<smbrandonjr-Re5JQ...@public.gmane.org> wrote:

>@Paul - Would you be able to take a close up picture of the backside
>solder points of the serial debug connector (3 pin next to USB C)?
>

Not "Paul" but... attached image I hope


>My AI was overheating too until it stopped booting all together. I was
>looking to monitor serial as I applied power and that is when I noticed
>what appeared to be an unclean solder point...its not shorted there is a
>99kOhm resistance between the two points.
>

Similar... Do I take a needle file to it, or risk a soldering gun... Or
complain to Newark about a defect...

--
Dennis L Bieber

solder_bridge.jpg

Theodore A. Roth

unread,
Sep 28, 2019, 1:48:30 PM9/28/19
to beagl...@googlegroups.com
Isn't that just a pull-up or pull-down resistor? My board has it too.


Ted Roth

--
For more options, visit http://beagleboard.org/discuss
---
You received this message because you are subscribed to the Google Groups "BeagleBoard" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beagleboard...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/beagleboard/ch0voeducvai6svmhn2nmleld0nf9di9ga%404ax.com.

Robert Forsyth

unread,
Sep 28, 2019, 2:17:48 PM9/28/19
to BeagleBoard
Theodore is correct, pull-down on debug serial RX, it is in the release notes https://github.com/beagleboard/beaglebone-ai/issues/24

I assumed the PCB layout had been changed.

On Thursday, 26 September 2019 00:51:41 UTC+1, Paul Richards wrote:

jks

unread,
Sep 28, 2019, 2:48:34 PM9/28/19
to BeagleBoard
Yes, a hand-soldered pulldown:

BBAI_serial_pulldown.JPG



Dennis Lee Bieber

unread,
Sep 28, 2019, 3:11:39 PM9/28/19
to Beagleboard
On Sat, 28 Sep 2019 11:48:34 -0700 (PDT), in
gmane.comp.hardware.beagleboard.user jks <jks-t5F...@public.gmane.org>
wrote:

>Yes, a hand-soldered pulldown:
>
That may have been the attempt, but looking at mine (see previous
attachment) the resistor has been completely covered with solder, and
solder also fills in the area under it down to the board level.
--
Dennis L Bieber

Jason

unread,
Oct 2, 2019, 7:24:39 AM10/2/19
to beagl...@googlegroups.com
On Sat, Sep 28, 2019 at 8:19 AM Mike Brandon <smbra...@gmail.com> wrote:
@Paul - Would you be able to take a close up picture of the backside  solder points of the serial debug connector (3 pin next to USB C)? 

My AI was overheating too until it stopped booting all together. I was looking to monitor serial as I applied power and that is when I noticed what appeared to be an unclean solder point...its not shorted there is a 99kOhm resistance between the two points.



Mike

On Wednesday, September 25, 2019 at 7:51:41 PM UTC-4, Paul Richards wrote:
I got 2 BeagleBone AI's from Mouser and started working with them today.

Both of my boards boot and run with the 08-03 build.

I haven't done anything to modify the contents of the filesystem, I merely booted and connected via Cloud9.

On both boards, the main CPU heat sink and the Kingston RAM module immediately
heat up and run incredibly hot.  I measured >80C on the CPU and >90C on the RAM module with my IR thermometer.

System load looked fine, nothing was running away and stealing all the CPU.

We intend to use these in lieu of BeagleBone Black modules for a project in the Keck telescopes and if they normally run this hot, it will never work.

S/N range is 4000167 1301 003491 1933


Is my experience atypical?  What temperature should these boards nominally run at?



Paul

--
For more options, visit http://beagleboard.org/discuss
---
You received this message because you are subscribed to the Google Groups "BeagleBoard" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beagleboard...@googlegroups.com.

hli...@gmail.com

unread,
Oct 10, 2019, 9:27:22 AM10/10/19
to BeagleBoard
Hi Paul.

I see the same. I'm using the base OS as shipped (Debian 2019-08-03). I upgraded Debian and I upgraded my kernel with the tool in /opt to get to version 4.14.108-ti-r119.

If I just let the machine idle at the desktop it will get hot and shut down after a few minutes.

We want to use this as a controller without active cooling.

Is this software and drivers? Is this hardware? Is there a reasonable expectation that we could run the board with high load without a fan?

Is there a way to throttle processor?


Thanks,
Harry

aras...@gmail.com

unread,
Oct 10, 2019, 9:28:55 AM10/10/19
to BeagleBoard
If you are running the device headless (no display attached) you can disable a few things that are display related in the device tree (HDMI, GPU). That will eliminate about 500mW of power consumption / heat. You should also disable lightdm (lxqtpanel process is constantly using 10% of the CPU). 

Finally, if you're fanless get a different heat sink! The included one is only meant for fans. This is the one I'm using: https://ebay.us/RUFBGy

With a better heat sink and disabling GPU my idle temperatures are 55C rather than 80C even with the cpu governor set to "ondemand". 

omapconf printout (with MPU at 1GHz minimum)

On Wednesday, September 25, 2019 at 6:51:41 PM UTC-5, Paul Richards wrote:
Reply all
Reply to author
Forward
0 new messages