Page 5 of 5 FirstFirst 12345
Results 41 to 50 of 50

Thread: Banana Pi

  1. #41
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    Quote Originally Posted by NeoGen View Post
    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
    I've looked into similair cases, but I've yet to find one where the cluster can share both WUs and memory (e.g. work on one WCG WU over 12 cores by using the RAM of the entire 40-board cluster)

    Quote Originally Posted by NeoGen View Post
    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.
    It is amazing to see how fast a question as "What would you do with a 120-Raspberry Pi Cluster?"
    can be translated into "What would you do with a 15-Banana Pi M3 Cluster?" and getting more performance for less wattage.

    Quote Originally Posted by NeoGen View Post
    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
    I hadn't yet, but they are small in both size (NanoPi-2 and NanoPi-2-Fire are hardly bigger than a Raspberry Pi Zero)
    and price (The quad-core Allwinner H3 NanoPi-M1 can be bought for a price as low as 13 US$, the octo-core Nanopi-M3 is,
    together with the NanoPi-2 and NanoPi-2-Fire, the most expensive at only 32 US$)
    The models M1, M2 and M3 all slightly differ in size (64 mm x 50-56-60 mm, but are generally smaller than the Raspberry Pi A+ (65 mm x 56.5 mm).
    Last edited by Dirk Broer; 07-14-2016 at 09:28 PM.


  2. #42
    AMDave's Avatar
    AMDave is offline Seeker of the exit clause Moderator
    Site Admin
    Join Date
    Jun 2004
    Location
    Deep in a while loop
    Posts
    9,609
    sounds more like a Beowulf Cluster, then.
    Try this:
    http://coen.boisestate.edu/ece/files...Cluster_v2.pdf
    and obviously further examples about:
    https://www.google.com.au/?client=ub...r+raspberry+pi
    but then you need to be compiling and running MPI coded applications to effectively run across the nodes. BOINC & BOINC apps will not do that for you as they are single threaded by design.
    There are a very few parallel apps but in most cases it would be recommended that the code be recompiled on the cluster to suit that cluster due to most Beowulf clusters being very different to each other.
    https://www.google.com.au/?client=ub...eowulf+cluster
    So you would only build one if you had planned ahead and are well prepared to write / re-compile apps for your cluster.
    That said there are many MPI schedulers and utilities available (instead of BOINC) for Beowulf clusters.
    I'm interested to see where you go with this.
    I'd more likely go with a network of individual devices, using a PXE host to manage the images so you can PXE boot the SBCs / TCs from the host then they don't each need a HDD and you can select whatever image you want as you boot them (with a temporarily attached screen) or preset with a script and monitor each remotely. But that's just me.
    I did build a 2 node Beowulf for fun and self-education about a decade ago (probably a lot longer - sshh!) with old hardware and got something parallel running on it, Folding / Distributed folding I think it was.
    Wow I even did post briefly about my attempt when I started the LTSP install - http://www.amdusers.com/forum/showth...wulf#post21911
    (11 years ago - Tempus fugit!) but not much after that
    I see in some PMs and archived posts that I was still having compatibility issues a year later and reworking it - url not available.
    But it was hard work and the machines got really hot and although the parallelism was interesting to look at, most of the time it did not use the resource of the whole cluster, about a 1/4 of it if I recall. YMMV.
    more chit-chat again in 2006 - http://www.amdusers.com/forum/showth...hlight=beowulf
    Of course the software and utilities and instructions are all much better and more available now and more widely in use.
    You can build a "render wall" very cheaply these days.
    Last edited by AMDave; 07-15-2016 at 12:21 AM.
    . . . . . ___
    . . . . . . .\___/\______
    . . . . . . . \__AMD___\\__
    ---------------------------------------------

  3. #43
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    So far the easiest solution -given the limitations of most BOINC applications- seems to be running BOINCTasks on a x86-64 machine and run the ARM boards -headless- through BOINCTasks.
    Last edited by Dirk Broer; 07-15-2016 at 12:55 AM.


  4. #44
    AMDave's Avatar
    AMDave is offline Seeker of the exit clause Moderator
    Site Admin
    Join Date
    Jun 2004
    Location
    Deep in a while loop
    Posts
    9,609
    Folding no longer support multi-node clusters using MPI
    http://folding.stanford.edu/home/faq/faq-smp
    Clients are now SMP which means running on a single machine with large numbers of cores / blades.
    The rational is that communication between cluster nodes is too slow.
    . . . . . ___
    . . . . . . .\___/\______
    . . . . . . . \__AMD___\\__
    ---------------------------------------------

  5. #45
    AMDave's Avatar
    AMDave is offline Seeker of the exit clause Moderator
    Site Admin
    Join Date
    Jun 2004
    Location
    Deep in a while loop
    Posts
    9,609
    There are other reasons for building a small cluster, like thus
    http://diybigdata.net/2016/06/instal...d-xu4-cluster/
    Last edited by AMDave; 07-15-2016 at 04:54 AM.
    . . . . . ___
    . . . . . . .\___/\______
    . . . . . . . \__AMD___\\__
    ---------------------------------------------

  6. #46
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    I finally seem to have my Banana Pro running stable under Ubuntu Mate 15.04 (no more tries to upgrade it), using an old 80486 heatsink (45x45x15 mm)

    that covers both the Allwinner A20 SOC and the RAM chips. Temps are 55 Celsius / 132 Fahrenheit / 328 Kelvin.

    I may have called the BeagleBone Black the total and utter king of blinkenlights before -due to the eternally flashing blue LED lights- but the Banana Pro comes as a close 2nd with a eternally burning red, a constantly flashing green and a less frequently flashing blue LED. Now for a casing that fits the relatively huge heatsink.....
    Last edited by Dirk Broer; 08-05-2016 at 01:44 AM.


  7. #47
    Join Date
    Nov 2005
    Location
    Central Pennsylvania
    Posts
    4,333
    Gentlemen this might help in your robust enthusiasm of Parallel computing: http://daugerresearch.com/vault/index.shtml





    Challenge me, or correct me, but don't ask me to die quietly.

    …Pursuit is always hard, capturing is really not the focus, it’s the hunt ...

  8. #48
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    Banana Pi have released yet another version of their quad-core M2, the M2 Ultra. What is the difference between these three M2 boards now?
    Feature/Board Banana Pi M2 Banana Pi M2+ Banana Pi M2 Ultra
    Size
    92mm x
    60mm
    65mm x
    65mm
    92mm x
    60mm
    SOC
    Allwinner
    A31s
    Allwinner
    H3
    Allwinner
    R40
    CPU
    ARM
    Cortex-A7
    ARM
    Cortex-A7
    ARM
    Cortex-A7
    Architecture
    ARMv7-A
    ARMv7-A
    ARMv7-A
    Speed
    1200 MHz
    1200 MHz
    ? MHz
    Cores
    4
    4
    4
    RAM
    1024 MB
    1024 MB
    2048 MB
    RAM/Core
    256 MB
    256 MB
    512 MB
    L2 Cache
    1024 KB
    512 KB
    ? KB
    GPU
    Dual core PowerVR SGX54 MP2 @350MHz
    Dual core Mali-400 MP2 @600MHz
    Dual core Mali-400 MP2 @600MHz
    USB 2.0
    4
    2
    3 (1 via GPIO header)
    LAN
    10/100/1000
    10/100/1000
    10/100/1000
    WiFi
    802.11b/g/n
    802.11b/g/n
    802.11b/g/n
    Bluetooth
    No
    4.0
    4.0
    SATA
    No
    No
    Yes
    eMMc
    No
    8GB
    8GB (standard)
    16/32/64GB (optional)
    So far too much '?' with the Ultra to be really able pointing out the best board for you, but the M2 Ultra does have some nice features.
    Last edited by Dirk Broer; 09-10-2016 at 07:46 PM.


  9. #49
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    The Banana Pi family differs hugely from that of the Raspberry Pi. The arch version (nowadays called Banana Pi BPI-M1) was April 2014 brought out as an attempt at a slightly better (and bigger) Raspberry competitor, offering a dual-core ARM Cortex-A7 CPU (when the raspberry still did with a ARM11), Gigabit ethernet, an IR sensor and a SATA connector, amongst others. It lacked however the communiuty, software and hardware support of the Raspberry.


    Within the first Banana Pi generation are two branches that appeared around the time Raspberry brought out their improved Model B+. The first to emerge was in October 2014 the LeMaker-produced Banana Pro, followed April 2015 by the Sinovoip-produced Banana Pi BPI-M1+. Both companies had produced the original Banana Pi BPI-M1 and another variant: the Banana Pi BPI-R1 (October 2014, basically a Banana Pi BPI-M1 with four additional ethernet ports and WiFi). The differences between the Banana Pi Pro and the Banana Pi BPI-M1+ are mainly in the colour of the PCB and the placement of the Allwinner A20 SOC. LeMaker has it at the same place as the original Banana Pi (at the rearside), Sinovoip aka Banana-Pi.Org having the SOC on top.

    But in 2015 Raspberry.org was developping the quad-core ARM Cortex-A7 CPU equipped Raspberry Pi 2B, causing Banana-Pi.org to responded in April 2015 with the likewise (Allwinner A31s) equipped Banana Pi BPI-M2 (since discontinued), and in November 2015 with the octo-core ARM Cortex-A7 (Allwinner A83t) equipped Banana Pi BPI-M3 and in November 2016 the Banana Pi BPI-M2 Ultra.

    Banana-Pi.org brings out a still increasing greater number of boards, such as in April 2016 the Raspberry Pi A+ like Banana Pi BPI-M2+ -where Banana-Pi.org again manages to offer much more features than the competing Raspberry product, such as a quad-core Cortex A7 (Allwinner H3), 8GB onboard eMMC, WiFi and BlueTooth and 10/100/1000Mbps Ethernet. BTW: the M2+ comes in H3, Edu, H2+ ad H5 variants, of which the last one, the H5 is a hybrid 2nd/3rd generation banana as it features a 64-bit quad-core Cortex-A53 SOC. For those than can do with even less features there's since November 2017 the Banana Pi BPI-M2M aka Banana Pi M2 Magic. Other raspberry-inspired Banana boards are the Banana Pi BPI-M2 Berry (brought out in May 2017, finally having the same size as the Raspberry Pi family) and the Banana Pi BPI-Zero, that packs a quad-core ARM Cortex-A7 on a board the size of a Raspberry Zero (where that board still has to do with a single-core ARM11). The Banana Pi BPI-M64, introduced in November 2016 was no doubt inspired by the ARM Cortex-A53 CPU equipped Raspberry Pi 3B, as it features a likewise Allwinner A64 or R18, depending on the sub-version.


    But Banana-Pi.org does quite original things too. They improved upon the Banana Pi BPI-R1 router with their 2017 Banana Pi BPI-R2, offering a MediaTek MT7623N (a Quad-core ARM Cortex-A7), Mali 450 MP4 GPU, 2G DDR3 SDRAM and a Mini PCIE interface. Another router-like product is the Banana PI BPI-W2. They also have since April 2015 the Banana Pi BPI-G1, which is a next generation of IoT hub with WiFi, Zigbee, and BT support. For those even deeper in the internet-of-Things (IoT) there's the Banana Pi BPI-D1, a SBC/camera combo


    LeMaker has stopped making Banana called products after the Banana Pro and their 2nd generation now has music-inspired names such as the vaguely Raspberry compute module-like LeMaker Guitar, the LeMaker Piano and the LeMaker HiKey. These boards not even remotely look Raspberry-like.
    Last edited by Dirk Broer; 01-22-2021 at 11:01 PM.


  10. #50
    Join Date
    Sep 2010
    Location
    Leiden, the Netherlands
    Posts
    4,372
    Sinovoip, producer of Banana Pi boards, has a new flag ship in their M (for main?) line, perhaps unsurprisingly called the Banana Pi M6 (don't get distracted by the header showing a M4).
    The surprise lies more in the chosen SOC: a Senary (Synaptics) VS680, a quad-core Cortex-A73 (@2.1GHz) with a Cortex-M3 real-time security core @ 250 MHz (-so useless for computing), an Imagination PowerVR Series9XE GE9920 GPU, and a NPU for AI up to 6.75 Tops.
    The board further has 4GB LPDDR4, 16GB eMMC flash and offers a M.2 E-Key for PCIe or MIPI CSI. There are 4 USB 3.0 ports, 1 GbE ethernet and 1 Micro HDMI-in and 1 Micro HDMI-out. Power Source needs to be PD(?) 5V@3A, administered via USB 3.0 Type-C.
    The board looks like this (click to enlarge):

    The VS680/SN3680 seems to be a quite capable SOC -but these specs make me want a SBC with a Qualcomm QCS605 instead.
    More info:
    https://www.cnx-software.com/2022/11...-ai-processor/
    https://linuxgizmos.com/banana-pi-in...680-based-sbc/


Page 5 of 5 FirstFirst 12345

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •