Banana Pi

[Brucifer]

Lol. Almost like someone on the design team actually plays with the product and is aware of conflicts that could arise!! :-)

[Nflight]

You clearly have my attention and have spent more time reading and updating myself with these posts. Nice Job Dirk, and I too am following along as you progressively display the nuances of the tiny computer. Very happy with what has been displayed so far, keep going! All eyes forward...

[Brucifer]

So is anyone using any of these to crunch with?? If so, what and what project(s)??

[Dirk Broer]

Chris has an active Banana Pi M2 quad running under Android at the moment, I am readying a Banana Pro dual-core.
For these ARM devices Android is the OS of choice, but I might try Raspbian for the Banana Pro too,
that would give me the chance to test the Pro against the Raspberry Pi2.

Projects known to work with the mini's include Asteroids, Enigma, Seti, Yoyo (OGR cruncher), Citizen Science Grid (SubsetSum), WCG, Einstein, MilkyWay.
Look at this posting for projects running on ARM using Android.

[Brucifer]

Thank you Dirk! :-)

[Dirk Broer]

I've got a raspberry pi but now I guess I'll have to get one of them there banana things. :-)

After the invasion of the Raspberry Pies there followed a rain of other fruit pies, the Banana Pi the most persistent so far.
Offered by two companies (BPI and Lemaker, who seems to have split up after the intial M1 and R1 models that they made -and still sell- together) they have rows of Banana's waiting for us.
The specs -at least- are good:

Feature	Banana Pi M1	Banana Pi M1+	Banana Pro	Banana Pi M2	Banana Pi M2+	Banana Pi M3	Banana Pi R1
Size	92mm x 60mm	92mm x 60mm	92mm x 60mm	92mm x 60mm	65mm x 65mm	92mm x 60mm	148mm x 100mm
SOC	Allwinner A20	Allwinner A20	Allwinner A20	Allwinner A31s	Allwinner H3	Allwinner A83t	Allwinner A20
CPU	ARM Cortex-A7	ARM Cortex-A7	ARM Cortex-A7	ARM Cortex-A7	ARM Cortex-A7	ARM Cortex-A7	ARM Cortex-A7
Architecture	ARMv7-A	ARMv7-A	ARMv7-A	ARMv7-A	ARMv7-A	ARMv7-A	ARMv7-A
Speed	1000 MHz	1000 MHz	1000 MHz	1200 MHz	1200 MHz	1800 MHz	1000 MHz
Cores	2	2	2	4	4	8	2
RAM	1024 MB	1024 MB	1024 MB	1024 MB	1024 MB	2048 MB	1024 MB
RAM/Core	512 MB	512 MB	512 MB	256 MB	256 MB	256 MB	512 MB
USB 2.0	2	2	2	4	2	2	2
LAN	10/100/1000	10/100/1000	10/100/1000	10/100/1000	10/100/1000	10/100/1000	4x 10/100/1000 LAN 1x 10/100/1000 WAN
WiFi	No	Yes	Yes	Yes	Yes	Yes	Yes
Bluetooth	No	No	No	No	Yes	Yes	No
SATA	1x SATA-150	1x SATA-300	1x SATA-300	No	No	1x SATA-300	1x SATA-150
eMMc	No	No	No	No	8GB	8GB	No

Despite being not as popular as the Raspberries, the Bananas do have some claim to fame.

• The original Banana Pi -now called the M1 and with two improved variants, Pro and M1+ - was a ARMv7 dual core with 1 GB RAM, when the original Raspberry Pi B still was a ARMv6 single core with 512 MB -and the original Raspberry Pi A still was a ARMv6 single core with 256 MB.
• The Banana Pi M2 already was a ARMv7 quad core -with higher clock speed- before the Raspberry Pi 2 arrived.
• The Banana Pi M3 is an ARMv7 octa-core, a feat which Raspberry has not managed -yet.
• The Banana Pi M2+ can be compared with the Raspberry Pi A+ (format-wise), but here the advantage is again with the Banana: it is a ARMv7 quad core with 1 GB RAM, while the Raspberry Pi A+ still is a ARMv6 single core with 256 MB.

Me personally, I am waiting for a Banana Pi R2, featuring all extra's (quad- or octo-core, WiFi, Bluetooth, eMMc, SATA-300).

[Dirk Broer]

My Banana Pro has come to life!

After my Raspberry Pi 2 -the one that runs Ubuntu Mate 16.04- suffered from a corrupted SD-card -most likely from either the screensaver or the rotating background picture, disable that feature!- and crashed, I decided to try the Banana Pro with a new, better SD card fitted with the latest Bananian. Now I am trying to get BOINC installed....even when Bananian is no sucess -as it looks like now- I can now install e.g. Raspbian, or Ubuntu Mate.

Edit 1: Right now it is updating from Ubuntu Mate 15.04 to 15.10....edit: from 15.10 to 16.04
Edit 2: That was not such a bright idea, as the kernel remains at 3.4.103 and gets out of sync with other software. Back to 15.04, and update the kernel first then.

When fitted with a bad image -any image made with the aid of the Allwinner utility PhoenixCard in my case- only a red light will shine, and you get no video. I was about to bin the board.
A good image -so far tried Bananian (a very minimalistic Debian) and Ubuntu Mate- gives red, blue and green lights when booting, and video afterwards.

I am impressed by the BOINC benchmarks for this ARM Cortex-A7 board, up to 565 floating point MIPS (Whetstone) per CPU and 1995 integer MIPS (Dhrystone) per CPU, but it is a pity there are only two such cores. Still, it kicks the s**t out of the -single core- Raspberry Pi B+ and Beaglebone Black and even has a better performance per core than the Pi 2 quad-core (Enigma in a little more than 3 hours!).

BTW, I am less impressed with the stability of Ubuntu Mate 16.04 for Raspberry -it keeps trashing the SD card- and the WiFi connection of Ubuntu Mate 16.04 for the Banana.
Once upgraded to 16.04 it tends to loose the WiFi connectivity.

[Dirk Broer]

What I ultimately want for my SBC boards is a set-and-forget way of operating. I want to power them up, install the OS and BOINC, point them to their projects and check on them once a month or so. This however was not possible with Ubuntu Mate on the Banana Pro, nice experience as it was to use a full desktop on an ARM development board though.

Next stop is OpenSuse -too old (13.1, won't update. Too much unneeded stuff as well),
then it is ArchLinux, this image went wrong because the ArchLinux servers seem unreachable right now, at least from the LeMaker supplied -old- image.
Up to Gentoo then, but no longer via a LeMaker image!

If that fails, it is back to Bananian...

[AMDave]

You could set up a PXE host (even in a VirtualBox) and PXE boot all oof your SBCs from the host
https://www.reddit.com/r/raspberry_p...ork_boot_boot/

That way you can put together various OS/App configs in a single easy to access place and try them out on all of your SBCs very quickly.
But since you want to set and forget, then just make them headless and install the server version without the desktop. BOINC mgr won't run, of course, but you would not need to, just run the client and RPC to it from one of your AMD workstations.
You will learn a lot by doing this.

Fedora/REDHAT/CentOS and Debian/Ubuntu LTSP server spins already exist almost fully configured (better configured than they used to be)
https://help.ubuntu.com/community/Ub...SPQuickInstall

Also there is another github project called FOG which allows you to d/l & install & run, however the built-in default PXE image is likely to be 32/64 intel/AMD so you'd have to go the extra step and create an ARM image for your SBCs which is what FOG is good at, it will quite happily allow you to build a Win## (any version) image and PXE that to your devices machine.
This is how many thin-client boxes are used in the work-place although the host configurations vary as there are many ways to serve PXE to the network.
https://www.google.com.au/?client=ub...p+a+pxe+server

Some folk are PXEing their XBMC to SBCs & TCs around the site so they have the same GUI in all locations.

EDIT - I see I mixed my technologies again. Better trying FOG first which is PXE of images over the LAN, where-as the linux LTSP editions are full LTSP over PXE and, while they can server images over PXE, they require quite a bit more work if you want to fully leverage the capabilities and benefits of LTSP, like using a single btrfs stored linux OS to server to all of your different hardware which is very fast to the clients and a massive reduction on disk space and RAM on the host. Or you could just go the old rout which I didn't mention which is to connect the PXE boot devices to server based sessions that run on the server and not on the device (real thin-client) but thats really for better desktop experience than the TC would normally provide. Lots of possibilities.

[NeoGen]

I was looking around on google for the possibility of building a little cluster out of those boards, and some people have done it indeed
http://www.owncluster.de/2015/07/08/...na-pi-cluster/
http://www.jeffgeerling.com/blog/201...rry-pi-cluster

That might be complex to achieve but interesting to play with, in the end it should be fully controlled as a single node but have dozens of cores and a large stack of RAM depending on how many of them you cluster together.


But then by accident I stumbled on a site with even smaller boards... have you seen the NanoPi's?
http://www.nanopi.org/index2.html