Beaglebone Black

[Dirk Broer]

The Beaglebone Black, in its Rev.C, is a worthy opponent for the Raspberry Pi B+.
In fact the Beaglebone has more than an edge on the single-core Raspberries (A, A+, B and B+)
by virtue of its 1000 MHz ARM Cortex-A8 (http://en.wikipedia.org/wiki/ARM_Cortex-A8) CPU, its 4GB onboard storage -and the USB cable that comes free with it.

Compared to the Raspberry you can save on the cost of an external SD-card and the USB cable,
and suddenly the price difference between the boads has -more than- evaporated.
http://robosavvy.com/store/images/beagleboneDesc.jpg
Rev.B, which still used 2GB onboard storage.

Doesn't the beaglebone Black have any draw-back? Well, it only has one USB2.0 port, but then again:
you can use keyboard, monitor and mouse of any Windows or Linux PC to control it over the supplied USB cable.
And just like the Raspberry Pi A versions, you can make use of a USB hub when you need more than one USB port.

And like the Raspberry Pi, the Beaglebone has much hardware made to fit: there are shields especially made for the Beaglebone,
just like for the Raspberry Pi. There is e.g. a Logi Bone -just as there is a Logi Pi- to turn your bord into a FPGA tool;
there are Beaglebone to Arduino shields, so your Beaglebone should be able to carry e.g. a radiation shield too, just like the Raspberry.

So far the only thing that really sets the Beaglebone Black aside from the Raspberry Pi is the lack of a quad core development, a Beaglebone Black2.

I'll go and test it, using BOINC, and let you know what fares better: The Raspberry Pi B+, the Beaglebone Black Rev.C, the Banana Pro or the Raspberry Pi2.

As it turns out the Raspberry Pi2 fares best, followed by the Beaglebone, The Raspberry Pi B+ and lastly the Banana Pro (least stable of the four, needs constant care/reformatting its SD-card).

[Dirk Broer]

By the way, the Beaglebone Black is the total and utter king of Blinkenlights (http://en.wikipedia.org/wiki/Blinkenlights) as compared to the other single boards I own.


ACHTUNG!
ALLES TURISTEN UND NONTEKNISCHEN LOOKENPEEPERS!
DAS KOMPUTERMASCHINE IST NICHT FÜR DER GEFINGERPOKEN UND MITTENGRABEN! ODERWISE IST EASY TO SCHNAPPEN DER SPRINGENWERK, BLOWENFUSEN UND POPPENCORKEN MIT SPITZENSPARKEN.
IST NICHT FÜR GEWERKEN BEI DUMMKOPFEN. DER RUBBERNECKEN SIGHTSEEREN KEEPEN DAS COTTONPICKEN HÄNDER IN DAS POCKETS MUSS.
ZO RELAXEN UND WATSCHEN DER BLINKENLICHTEN.

[Nflight]

Your descriptions are always spot on descriptive. I love the extensive descriptions leaving nothing to hide and explaining everything I would need to know!

[Dirk Broer]

Running straight from the onboard Ångström Linux did not quite work out:

root@beaglebone:/var/lib/boinc-client# boinc
31-Mar-2015 21:57:48 [---] Starting BOINC client version 7.0.27 for arm-unknown-linux-gnueabihf
31-Mar-2015 21:57:48 [---] log flags: file_xfer, sched_ops, task
31-Mar-2015 21:57:48 [---] Libraries: libcurl/7.26.0 OpenSSL/1.0.1e zlib/1.2.7 libidn/1.25 libssh2/1.4.2 librtmp/2.3
31-Mar-2015 21:57:48 [---] Data directory: /var/lib/boinc-client
31-Mar-2015 21:57:48 [---] Processor: 1 ARMv7 Processor rev 2 (v7l)
31-Mar-2015 21:57:48 [---] Processor features:
31-Mar-2015 21:57:48 [---] OS: Linux: 3.8.13-bone70
31-Mar-2015 21:57:48 [---] Memory: 496.47 MB physical, 0 bytes virtual
31-Mar-2015 21:57:48 [---] Disk: 3.44 GB total, 1.43 GB free
31-Mar-2015 21:57:48 [---] Local time is UTC +0 hours
31-Mar-2015 21:57:48 [---] No usable GPUs found
31-Mar-2015 21:57:48 [---] Config: GUI RPC allowed from:
31-Mar-2015 21:57:48 [SETI@home] URL http://setiathome.berkeley.edu/; Computer ID 7533940; resource share 10
31-Mar-2015 21:57:48 [---] No general preferences found - using defaults
31-Mar-2015 21:57:48 [---] Reading preferences override file
31-Mar-2015 21:57:48 [---] Preferences:
31-Mar-2015 21:57:48 [---] max memory usage when active: 248.24MB
31-Mar-2015 21:57:48 [---] max memory usage when idle: 446.83MB
31-Mar-2015 21:57:48 [---] max disk usage: 1.34GB
31-Mar-2015 21:57:48 [---] don't use GPU while active
31-Mar-2015 21:57:48 [---] suspend work if non-BOINC CPU load exceeds 25 %
31-Mar-2015 21:57:48 [---] (to change preferences, visit the web site of an attached project, or select Preferences in the Manager)
dir_open: Could not open directory 'slots'.
31-Mar-2015 21:57:48 [---] Not using a proxy
Initialization completed
31-Mar-2015 21:57:49 [---] Running CPU benchmarks
31-Mar-2015 21:57:49 [---] Suspending computation - CPU benchmarks in progress
31-Mar-2015 21:58:20 [---] Benchmark results:
31-Mar-2015 21:58:20 [---] Number of CPUs: 1
31-Mar-2015 21:58:20 [---] 184 floating point MIPS (Whetstone) per CPU
31-Mar-2015 21:58:20 [---] 2047 integer MIPS (Dhrystone) per CPU


Note the small value for floating point MIPS, I guess it does not quite make use of the NEON capabilities.
Nothing wrong with the integer value -as compared to the original Raspberries- though!


31-Mar-2015 22:16:57 (low) [SETI@home] update requested by user
31-Mar-2015 22:16:59 (low) [SETI@home] Sending scheduler request: Requested by user.
31-Mar-2015 22:16:59 (low) [SETI@home] Not reporting or requesting tasks
31-Mar-2015 22:17:02 (low) [SETI@home] Scheduler request completed
31-Mar-2015 22:17:02 (user notification) [SETI@home] Message from server: This project doesn't support computers of type arm-unknown-linux-gnueabihf

I guess I will have to install Android on the BBB (=Beaglebone Black) in order to get work for the ARM Cortex-A8,
as I do not like the prospect of having to compile client or applications myself each time I need a new one...I've read some pretty positive feedback on NativeBoinc :icon_wink:

To be continued!

[Nflight]

By the way, the Beaglebone Black is the total and utter king of Blinkenlights (http://en.wikipedia.org/wiki/Blinkenlights) as compared to the other single boards I own.

Thanks for the History lesson on Blinkenlights, I always wanted to know where that descriptive nonsense came from ! :allhail:

[Dirk Broer]

When you want to use the Beaglebone Black for BOINC there are a few things to consider.


The CPU is a Texas Instruments Sitara AM335x, which is for all practical purposes an ARM Cortex-A8 core (http://www.arm.com/products/processors/cortex-a/cortex-a8.php).

(x depending on the model/revision. In the case of the Beaglebone Black Rev.C it is a Sitara AM3358BZCZ100).
Not all OS-es run on an ARM Cortex-A8, and those who do are not are not all supported by BOINC -at least not officially.

You may have to compile the client yourself (http://boinc.berkeley.edu/trac/wiki/BuildSystem). Do not forget the set the options specific for an ARM Cortex-A8:

CFLAGS="-march=armv7-a -mtune=cortex-a8" and

CFLAGS="$CFLAGS -mfloat-abi=hard -mfpu=neon -ffast-math -O3"
Having built a BOINC client yourself on your OS of choice does not guarantee that you can download work from a project -at least not automatically.

You may have to use the anonymous platform option (http://boinc.berkeley.edu/wiki/anonymous_platform).


The standard OS for the Beaglebone is the very lightweight Ångström Linux (http://www.angstrom-distribution.org/) and to my astonishment it does not make full use of the NEON floating point capabilities, giving the BBB a very low value for floating point MIPS as obtained through the benchmarks in the BOINC-client (184 Floating Point MIPS, see below).

Let's look at the BOINC projects that offer applications for the ARM Cortex-A8, so you do not have to compile them yourself (If you do, do not forget to compile the applications with the same CFLAGS values as mentioned for the BOINC-client!)



Project
Supported CPU architecture


ABC@Home
Only official support for x86, x86-64 and PPC (http://abcathome.com/apps.php)


Albert@Home
Due to the setup of their new Drupal site it is not feasible to do a lookup which CPU architectures are supported. So much for progress and inventing the wheel again. Native BOINC keeps stats for Albert (http://nativeboinc.org/site/host_stats#project=albert) though, so it runs under Android


AlmereGrid
x86, x86-64 only (http://boinc.almeregrid.nl/apps) and such recent builds too...


AlmereGrid TestGrid
x86, x86-64 only too I guess, but they are offline at the moment


Asteroids@Home
Should be possible using Android (http://asteroidsathome.net/boinc/apps.php)


Atlas@Home
x86-64 paradise (http://atlasathome.cern.ch/apps.php)


Bitcoin Utopia
Would need Raspbian and enabled Hardware Floating Point (http://www.bitcoinutopia.net/bitcoinutopia/apps.php), and that's a tall order for a Beaglebone Black.
There is a MineNinja that uses the Beaglebone Black, but I do not know as yet if it would work on BU


BOINC@Fiit
x86, x86-64 only (http://boinc.fiit.stuba.sk/apps.php)


BURP
You are invited to try for yourself (http://burp.renderfarming.net/forum_thread.php?id=2439&sort=5)


CAS@Home
x86, x86-64 only (http://casathome.ihep.ac.cn/apps)


Citizen Science Grid
x86, x86-64 only (http://volunteer.cs.und.edu/csg/apps.php)


Climateprediction.net
x86, x86-64 only (http://climateapps2.oerc.ox.ac.uk/cpdnboinc/apps.php)


Climate@Home
Offline


Collatz Conjecture
Mini Collatz and Micro Collatz can be run on ARM, using Android (http://boinc.thesonntags.com/collatz/apps.php)


Constellation
Automatically a x86 and x86-64 paradise (http://aerospaceresearch.net/constellation/apps.php), manually they extend to SPARC64, Itanium (IA64) and PARISC 2.0 (http://aerospaceresearch.net/constellation/apps.html)


Convector
Offline


Cosmology@Home
x86 and x86-64 paradise (http://www.cosmologyathome.org/apps.php)


DistributedDataMining Project
x86 and x86-64 paradise (http://www.distributeddatamining.org/ddm_apps)


DistributedRainbowTableGenerator (DistrRTgen)
Offline


EDGeS@Home
x86 and x86-64 paradise (http://home.edges-grid.eu/home/apps.php)


Einstein@Home
Mostly a x86 and x86-64 paradise (with some PPC), but the Binary Radio Pulsar Search (Arecibo) application has an ARM version running on Android (http://einstein.phys.uwm.edu/apps.php)


Enigma@Home
The most common architecture/OS combinations have almost all an executable here (http://www.enigmaathome.net/apps.php)


FightNeglectedDiseases@Home aka FiND@Home
(formerly FightMalaria@Home)
Besides the usual support for x86 and x86-64, they have recently added support for Raspberry Pi's (http://findah.ucd.ie/apps.php)
No mentioning of more general ARM support though


Gerasim
x86 and x86-64 only, they recently added GPU applications too (http://gerasim.boinc.ru/users/viewApps.aspx)


GPUGRID.net
There is an Android ARM32 application (https://www.gpugrid.net/apps).


Ibercivis
Yet another x86 and x86-64 paradise (http://registro.ibercivis.es/apps.php)


iGEM@Home
Offline


La Red de Atrapa Sismos
Should be possible under Android (http://www.ras.unam.mx/sensor/apps.php)


La Red de Atrapa Sismos Continuo
Should be possible under Android (http://www.ras.unam.mx/continual/apps.php)


The Lattice Project
Yet another x86 and x86-64 paradise (http://boinc.umiacs.umd.edu/apps.php)


Leiden Classical
x86, x86-64 only (http://boinc.gorlaeus.net/apps.php)


LHC@Home 1.0
SixTrack is x86 and x86-64 only (http://lhcathomeclassic.cern.ch/sixtrack/apps.php)


Malaria Control
All apps are x86 and x86-64 only (http://www.malariacontrol.net/apps.php)


MilkyWay@Home
Though not mentioned on the project site (http://milkyway.cs.rpi.edu/milkyway/apps.php), there is a Android application running through Native Boinc (http://nativeboinc.org/site/uncat/downloads)


MindModeling@Home (Beta)
So many applications, but yet another x86 and x86-64 whitewash (http://mindmodeling.org/apps.php)


Moo! Wrapper
One application is for Android running on ARM armeabi-v7a (http://moowrap.net/apps.php)


Najmanovich Research Group (NRG)
All apps are x86 or x86-64 only (http://boinc.med.usherbrooke.ca/nrg/apps.php)


NFS@Home
Despite the native FreeBSD support on all apps, they are x86 or x86-64 only (http://escatter11.fullerton.edu/nfs/apps.php)


NumberFields@Home
Another x86/x86-64 only project (http://numberfields.asu.edu/NumberFields/apps.php)


OProject@Home
Offline, but mentioned in the Native Boinc project statistics (http://nativeboinc.org/site/host_stats#project=oproject)


Optima@Home
All apps are x86 or x86-64 only (http://boinc.isa.ru/dcsdg/apps.php)


Pirates@Home
All apps are x86 or Mac OSX (intel or PPC) only (http://pirates.spy-hill.net/apps.php)


Poem@Home
All apps are x86 or x86-64 only (http://boinc.fzk.de/poem/apps.php)


Primaboinca
Offline


PrimeGrid
Though not mentioned on the project site (http://www.primegrid.com/apps.php), there is a Android application running through Native Boinc (http://nativeboinc.org/site/uncat/downloads)


QMC@Home
Though having x86 and x86-64 apps available for BOINC (http://qmcathome.org/apps), this project does not export stats -despite several requests- and is therefore to be avoided


Quake-Catcher Network Sensor Monitoring
Should be possible using Android (http://qcn.stanford.edu/sensor/apps.php), there are also Native Boinc Stats for QCN (http://nativeboinc.org/site/host_stats#project=qcn&offset=0&order=credits)


Quake-Catcher Network EMSC / CSEM
Should be possible using Android (http://qcn.emsc-csem.org/sensor/apps.php)


Quake-Catcher Network (Taiwan)
Should be possible using Android (http://qcn.twgrid.org/sensor/apps.php)

Radioactive@Home
Native BOINC has an ARM Android application to download (http://nativeboinc.org/site/uncat/downloads), but as yet no stats for this project
[/tr]

RALPH@Home
Rosetta Mini for Android runs on Android for ARM (http://ralph.bakerlab.org/apps.php)


Renderfarm.fi
Offline


Rioja Science
x86 and x86-64 only (http://boinc.riojascience.com/apps.php)


RNA World (beta)
Another x86/x86-64 whitewash (http://www.rnaworld.de/rnaworld/apps.php)


Rosetta@Home
Rosetta Mini for Android runs on ARM CPUs (http://boinc.bakerlab.org/rosetta/apps.php)


SAT@Home
x86/x86-64 only (http://sat.isa.ru/pdsat/apps.php)


Seti@Home
Will run on Android for ARM as there are stats -and downloads- at Native Boinc. (http://nativeboinc.org/site/host_stats#project=seti&offset=0&order=credits) This gets supported for SETI@home v7 with 5 application for ARM/Android on the Seti apps page (http://setiathome.berkeley.edu/apps.php)


Seti@Home Beta
Will run on Android for ARM as there are stats at Native Boinc. May need some fiddling though for a Cortex-A8 due to it having Neon, but VFPv3. You might need a ARMv6 client for the Beaglebone Black (http://nativeboinc.org/site/host_stats#project=setibeta&offset=0&order=credits)


theSkyNet POGS
Besides running on the usual x86/x86-64 CPU/OS combinations, fitsedwrapper also runs on Android running on ARM according to the website (http://pogs.theskynet.org/pogs/apps.php)


SLinCA
x86 heaven, 32 bits even (http://dg.imp.kiev.ua/slinca/apps.php)


SRBase
A x86/x86-64 paradise (http://srbase.myfirewall.org/sr5/apps.php)


SZTAKI Desktop Grid
Another x86/x86-64 paradise (http://szdg.lpds.sztaki.hu/szdg/apps.php)


TN-Grid
x86/x86-64 only (http://gene.disi.unitn.it/test/apps.php)


Universe@Home
x86/x86-64 paradise (http://universeathome.pl/universe/apps.php)


Universe@Home Test
x86/x86-64 paradise being tested (http://universeathometest.info/universe/apps.php)


Van Der Waerden Numbers
Another x86/x86-64 paradise (http://www.vdwnumbers.org/apps.php)


Virtual LHC@Home
Yet another x86/x86-64 whitewash (http://lhcathome2.cern.ch/vLHCathome/apps.php)


Volpex@Home
Yet another x86/x86-64 paradise (http://volpexathome.cs.uh.edu/VolPEx/apps.php)


WEP-M+2 (wanless2)
Another x86/x86-64 paradise (http://bearnol.is-a-geek.com/wanless2/apps.php)


World Community Grid
Judged from these stats it should be possible using Android through Native Boinc (http://nativeboinc.org/site/host_stats#project=wcg&offset=0&order=credits) and it is indeed: I just completed a FAAH WU on my Beaglebone Black.


WUProp@Home
All major architectures can be found here, both Datacollect 4 and 5 run ARM Android, 4 even ARM Linux (http://wuprop.boinc-af.org/apps.php)


Yafu
There is a testbenchmark running for Android running on ARM armeabi-v7a (http://yafu.myfirewall.org/yafu/apps.php)


Yoyo@Home
Cruncher ogr should be possible on an ARM Cortex-A8, both with Linux and Android (http://www.rechenkraft.net/yoyo/apps.php)

[Dirk Broer]

There is no official support for Android on Pi and the B+, being a ARMv6 -architecture- ARM11 -model-, is not very well suited for it too. Raspbian is just fine for it.
I reserve Android for my Banana Pi's and the Beaglebone Black -though I am slowly going nuts by the latter's refusal to boot from SD card.

The bone may have been driving me nuts, but I still got me marbles!
I appears that though the SD card is supposed to flash the eMMC (http://elinux.org/Beagleboard:Android), it is still needed at boot, and you need to connect the bone to a power source other than your PC, otherwise you still have the Debian that is on the eMMC (the 4GB internal flash storage of the BBB Rev. C).

First I connected the bone to my USB-powered hub -I also have an unpowered one which I use the expand the meagre one USB2.0 host port of the Bone.
So power comes in through the Micro USB client and leaves via the normal sized USB2.0 host to a 7-port USB2.0 hub. This I connect to a mouse and keyboard.
I have a micro HDMI to HDMI cable that connects with a HDMI Switch, so all my single boards can use the same monitor, that I connect through a HDMI to DVI adapter.

At first, using this setup, I got the full Beagle Debian -better than connecting the Bone to a Windows or Linux box, but still no cigar.
Then I decided to use the SD card that I had used to flash the bone.
Bingo! Android started!
Then I installed Native Boinc (http://nativeboinc.org/site/for_users/install), connected to BAM!... and now it's running....

http://xdec.de/wp-content/uploads/2014/08/native_boinc.png
It looks a bit like this, only my taskbar has less items, just the three on the left.

The different OS/Boinc Client combination has its effect: The Floating Point performance goes up to 277 MIPS (from 184, some 50%!), the Integer Performances drops to 1607 MIPS (from 2047, so about 20%). As it is still Boinc 7.0.36 there is no way of seeing the processor features, but nativeboinc supplies it through host info: swp, half, thumb, fastmult, vfp, edsp, thumbee, neon, vfpv3, tls.

[Nflight]

Really Dirk: "The bone may have been driving me nuts, but I still got me marbles! " Are you sure? :icon_lol: :blob3:

[Dirk Broer]

Who says I can't make witty remarks in another language?

[NeoGen]

I just happened to click and search through Native Boinc's page and noticed it has stats, teams, users, projects, etc... are those stats collected for work processed with the native boinc client only, or with any ARM capable boinc client?

[Dirk Broer]

I honestly don't know -though it appears to be so, there are not that much ARM clients around! And there were even less in the past...
BTW: The stats are not up-to-date either, when my Beaglebone is out of work I'll move to Boinc for Android, the official -and later released- client.

Which proves impossible as the official client reports 'no device' when I try to install it.

[Dirk Broer]

So smooth as the Raspberry can be brought to function in BOINC, so hard it was to hammer the bone in business.
It started with my discovery that running BOINC on an ARMv7 architecture CPU under Linux is no piece of cake. Most projects do NOT have an application, you have to compile them yourself. This I circumvented by first installing Android 4.4.4 and then installing Native BOINC.

I made a host group in BAM! called 'ARM Cortex-A8 on Android' and when starting Native BOINC first I selected BAM! as account manager.
This made all default applications come in. Lesson learned: do not make too many projects default within BAM! You might end up getting more than you want...

What is in the group?

Beaglebone Black, Rev. C, running Android 4.4.4 on a 3.18.5 kernel
CPU : A Texas Instruments ARMv7-A compatible processor ARM Cortex-A8, rev 2 (v7l) aka Sitara AM3358BZCZ100
Features : swp (http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dht0008a/CJHBGBBJ.html), half (https://en.wikipedia.org/wiki/Half-precision_floating-point_format), thumb (https://en.wikipedia.org/wiki/ARM_architecture#Thumb), fastmult (http://comments.gmane.org/gmane.linux.ports.arm.kernel/4578), VFP (https://en.wikipedia.org/wiki/ARM_architecture#Floating-point_.28VFP.29), edsp (http://www.arm.com/products/processors/technologies/dsp-simd.php), ThumbEE (https://en.wikipedia.org/wiki/ARM_architecture#Thumb_Execution_Environment_.28Th umbEE.29), NEON (https://en.wikipedia.org/wiki/ARM_architecture#Advanced_SIMD_.28NEON.29), VFPv3 (https://en.wikipedia.org/wiki/ARM_architecture#Floating-point_.28VFP.29), tls (https://en.wikipedia.org/wiki/Thread-local_storage)


My Beaglebone Black works standard on 1000 MHz and has an even prettier Lego casing than my Raspberry. 345It is the grey-black, 2nd to the left (click to enlarge).



Speed in MHz
OS
Floating Point MIPS (Whetstone) per CPU
Integer MIPS (Dhrystone) per CPU



1000

Beagle Debian Wheezy

184

2047



1000

Android 4.4.4

277

1607



1000

Beagle Debian Jessie

173

2173


It is stronger than the Raspberry B+ on integer MIPS performance, but a little bit behind in the floating point MIPS.
How does that translate into real-life performance?



Project
Avg. running time in hours
Avg. credit
Green light?
Orange light?
Red light?


Albert@Home

38

123

Yes

-

-


Asteroids@Home

tbd, first WU seems lost

tbd

Yes

-

-


Collatz@Home

tbd

tbd

tbd

tbd

tbd


Einstein@Home

44

63

Yes

-

-


Enigma@Home

3.1

45

Yes

-

-


GPUGrid

tbd

tbd

tbd

tbd

tbd


Moo! Wrapper

tbd

tbd

tbd

tbd

tbd


OProject@Home

tbd

tbd

tbd

tbd

tbd


PrimeGrid

tbd

tbd

tbd

tbd

tbd


QCN (Quake Catcher Network)

tbd

tbd

Yes*

-

-


Radioactive@Home

tbd

tbd

-

-

No*


Ralph@Home

tbd

tbd

tbd

tbd

tbd


Rosetta@Home

tbd

tbd

tbd

tbd

tbd


Seti@Home

tbd

tbd

tbd

tbd

tbd


Seti@Home Beta

tbd

tbd

Yes

-

-


The SkyNet POGS

125

361

Yes

-

-


World Community Grid

11.6 (FAAH)
4.2 (OET)

-

Yes

-

-


WUProp

6

14

Yes

-

-


Yoyo@Home

1.7

10.31

Yes

-

-


[]
Legenda

tbd=To be determined
*=Would need a sensor

- =
no data/not applicable

Green light=
Works out-of-the-box

Orange light=
Needs some fiddling

Red light=
Refuses to work



Soon to be updated too.

[Nflight]

Thanks for all your hard work on these showcasing the structure and ability of getting these all to work on multiple OS's & multiple projects. This must be a chore but you seem to appreciate the effort too. :)

[Dirk Broer]

Yesterday I powered up my Beaglebone Black using an undocumented feature.
As the number of SBCs I own is now greater than the number of ports (4) on my 1st HDMI hub I bought a 2nd -5-port- HDMI hub and devided the SBCs over the two hubs.
As the Beaglebone Black is painfully lacking in the USB-port department, I connected a powered USB 2.0 hub to the Bone to be able to use both mouse and keyboard.
I did so before I had connected the power. So the only ports in use were the micro HDMI (going to my monitor) and the USB host (going to the powered USB hub).
The USB client, which I normally use for powering-up the board, was still unconnected.
http://www.seeedstudio.com/document/pics/beaglebone.png
To my surprise I got video -Beagle Debian- and my surprise was even bigger when the video disappeared when I applied power, only coming back when I disconnected the USB client.
The board has more surprises, such as refusing to flash the eMMC -at least in the official way...

Though the BeagleBone Black is getting a bit long in the teeth, there are various efforts to prolong its useful life.
Among the better efforts is the SanCloud BeagleBone Enhanced
http://www.sancloud.co.uk/wp-content/uploads/2016/12/BBE.png
This has
1 GB RAM, as compared to the 512 MB of the BeagleBone Black, the BeagleBone Industrial (a.k.a. BeagleBone Red) or SeeedStudio's BeagleBone Green
1 mini-USB 2.0 client port and 4 USB 2.0 Ports, as compared to the 1 mini-USB 2.0 client port plus one USB 2.0 host port of the original BBB/BBR and the 1 micro-USB 2.0 client port plus 1 USB 2.0 host port of the BeagleBone Green.
Barometer, Accelerometer, Gyro and Temperature sensors (optional)
https://elinux.org/images/3/3f/SC-table.png

...but still the anemic single-core 1000 MHz AM3358 ARM Cortex-A8, instead of a quad-core A15 or A17. Even a dual- or quad-core Cortex-A9 or A7 would have been better.

[Dirk Broer]

The latest BeagleBone Black now has Bluetooth and WiFi, but still that infernal ARM Cortex-A8:
https://beagleboard.org/static/images/BeagleBoneBlackWireless_A3_horiz.jpeg

[Dirk Broer]

The latest BeagleBone Black now has Bluetooth and WiFi, but still that infernal ARM Cortex-A8

They've heard me, there in Texas:
https://beagleboard.org/static/images/BeagleBone_AI_angle_300.jpg
BeagleBone-AI preliminary specifications:

SoC: TI AM5729 dual core Cortex-A15 processor featuring 4 PRUs, Dual core C66x DSP, and 4 EVEs
System Memory: 1GB RAM
Storage: 16GB on-board eMMC flash with high-speed interface
Networking: Gigabit Ethernet and high-speed WiFi
USB: 1x USB type-C for power and superspeed dual-role controller, 1x USB type-A host
Expansion: BeagleBone Black (BBB) compatible headers
Dimensions: 86.4 x 53.4 mm (compatible with BBB)

[NeoGen]

Victory! 3 years later they finally did it! :icon_lol:

[Dirk Broer]

Victory! 3 years later they finally did it! :icon_lol:

And it is more than just a dual-core Cortex-A15, it has two dual core Cortex-M4's too, plus two FPUs (C66x), and a respectable GPU.
You may have to compile a special boinc-client, and special boinc apps too...

[NeoGen]

I did a quick search on those C66x FPUs, are they talking about the Texas Instruments C66x DSPs from 2011? - http://www.ti.com/lit/ml/sprt577b/sprt577b.pdf
If they are they seem to have some pretty interesting specs, but like you I also bet at least project apps would have to recompile their code to be able to use them.

[Dirk Broer]

It looks like the BeagleBone AI gets its AI from the two Dual-core ARM Cortex-M4's plus the two dual-core C66x DSPs.
Problem is getting those extra CPUs getting recognised and to work. CPU info for the BeagleBone AI only shows the dual-core Cortex-A15:


$ lscpu
Architecture: armv7l
Byte Order: Little Endian
CPU(s): 2
On-line CPU(s) list: 0,1
Thread(s) per core: 1
Core(s) per socket: 2
Socket(s): 1
Model: 2
Model name: ARMv7 Processor rev 2 (v7l)
CPU max MHz: 1500.0000
CPU min MHz: 1000.0000
BogoMIPS: 12.29
Flags: half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm

$ cat /proc/cpuinfo
processor : 0
model name : ARMv7 Processor rev 2 (v7l)
BogoMIPS : 12.29
Features : half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm
CPU implementer : 0x41
CPU architecture : 7
CPU variant : 0x2
CPU part : 0xc0f
CPU revision : 2

processor : 1
model name : ARMv7 Processor rev 2 (v7l)
BogoMIPS : 12.29
Features : half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm
CPU implementer : 0x41
CPU architecture : 7
CPU variant : 0x2
CPU part : 0xc0f
CPU revision : 2

Hardware : Generic DRA74X (Flattened Device Tree)
Revision : 0000
Serial : 0901700a7b180b22


ARM itself just released a new Cortex-M, the Cortex-M55, which they claim to be the most AI-capable Cortex-M processor and that delivers up to 5x DSP performance. https://images.anandtech.com/doci/15494/Arm-N55-6_575px.png
Its capabilities can be enhanced by coupling it to their Ethos-U55 NPU
-which can also be coupled to the Cortex-M33, M4 -like in the BeagleBone AI- and M7.
https://images.anandtech.com/doci/15494/Arm-N55-9_575px.png
https://images.anandtech.com/doci/15494/Arm-N55-10_575px.png
This might be useful for e.g. the Wildlife@Home sub-poject of Citizen Science Grid
-provided they have the energy to wake themselves up and write a new application.

[Nflight]

Ah wildlife@home, just how much video watching of nothing can you handle?

[Dirk Broer]

Ah wildlife@home, just how much video watching of nothing can you handle?

Me? None, that's where the AI comes in. AI is tireless...

[Dirk Broer]

They must have come to the conclusion that even a dual-core Cortex-A15 doesn't cut it anymore in the present AI (Artificial Intelligence) and ML (Machine Learning) days, so now they have come with
The BeagleBone AI-64 (https://beagleboard.org/ai-64)
https://www.marutsu.co.jp/contents/shop/marutsu/dgimg/goods/Photos/SeeedTechnologyLtd/MFG_102110646.jpghttps://forums.ghielectronics.com/uploads/db3298/original/2X/2/2094d15ac52076c21d3c08527e07afae22baaf3a.jpeg

Key features of the BeagleBone AI-64

Based on a powerful Texas Instruments TDA4VM processor with dual Arm Cortex-A72 microprocessor subsystem @ 2GHz
C7x+MMA (a matrix multiply accelerator of no less than 8 TOPS!) and 2 C66x floating-point VLIW DSPs
3x dual Arm Cortex-R5 co-processors
2x 6-core Programmable Real-Time Unit and Industrial Communication SubSystem (PRU-ICSSG)
4GB LPDDR4 RAM
16GB onboard eMMC flash storage
5V DC input power
USB 3.0 Type-C interface for power input and data
Dual USB super-speed (5Gbps) Type-A host ports
miniDisplayPort, 2 x 4-Lane DSI display connector
16-pin microcontroller header
Expansion headers compatible with many BeagleBone Black cape add-on boards
Gigabit ethernet M.2 E-key connector with PCIe, USB and SDIO for WiFi/Bluetooth and expansion
Dual 4-Lane CSI camera connectors
Micro-SD slot
1x Boot button, 1x Reset Button, 1x Power button
1x Power indication LED, 5x User LEDs
Wake-up domain serial port
Main domain serial port
JTAG 10pin Tag-Connector for debug
Assembled heat-sink


All this for some $180, I'm afraid...

[Dirk Broer]

I managed to find a 2nd hand BeagleBone AI on 'Marktplaats', the Dutch version of eBay, mixed with Craigslist.
The least it can do is crunch Universe@Home (https://universeathome.pl/universe/apps.php)....and I hope more.