We work on A64-OLinuXino the first Open Source Hardware 64-bit development board


A64 was announced January this year by Allwinner. It’s a quad core 64 bit Cortex-A53 processor targeting tablets.

Some features are:

  • H.265/H.264 video decoding in hardware
  • HDMI 4K display support
  • DDR3L memory with lower consumption is supported
  • eMMC 5.0 support

Last week we got notice that Allwinner have already A64 in production and we will get some chips with the software SDK for Android 5.1 based on Linux Kernel 3.10 by the end of the month (yes they finally escaped from Kernel 3.4 🙂 ), so we started to work on our A64-OLinuXino.

This chip is with humble peripherials, as Allwinner target and announcement is to sell it for $5 in volume (obviously not for mortals as we have to buy the first chips at higher than this price).

  • LCD parallel RGB, LCD LVDS, MIPI and HDMI video output
  • Camera interface
  • One USB Host
  • One USB-OTG
  • Two SPI
  • Four I2C
  • Six UARTs

As you can see no Ethernet, no SATA, no PCIe, but this is what you get for $5 right ?

What we found is also that although being 64-bit ARM the A64 memory bus is just 32-bit wide, we guess to make BOM smaller.

This will hurt the performance, so we have to see how it will behave compared to A20 and H3 for instance.

A64-OLinuXino initial specs will be:

  • A64 4 core Cortex-A53
  • AXP803 PMU + Audio support
  • 1GB or 2GB DDR RAM
  • 4GB NAND Flash oe eMMC
  • micro SD card connector
  • micro USB-OTG
  • 2Mpix Camera on board
  • HDMI 4K output
  • Audio Output, Input
  • LCD connector to connect with all our LCD-OLinuXino 4.3″ 7.0″ 10.1″ 15.6″
  • RTL8723BS WiFi + BT4.0 on board for connectivity
  • GPIO connector with access to all processor resources
  • 5V power input

This will be our first 64 bit ARM board and also the first OLinuXino we make completely with KiCAD.

If everything goes smoothly we will have first prototypes by end of November and depend on how fast we can test the software we may have these run in production for Christmas 🙂 the goal is to make 64-bit ARM with EUR 64 retail price!

As I expect A64 with 4 cores Cortex-A53 will behave well running Desktop Linux so we are evaluating the option to find some Chinese laptop manufacturer and ask them to supply us just with some light laptop body with decent 10.1-11.6″ LCD display, camera and keyboard with touch-pad. We can re-design A64 board to fit inside this laptop body and to make light under 1 kg laptop running Linux.

One of the problem is that A64 is quite new and no any Linux-Sunxi support, as nobody have seen A64 development boards yet. SO it may pass several months until A64-OLinuXino run anything else than Android 5.1

48 Comments (+add yours?)

  1. progmetalbg
    Oct 16, 2015 @ 17:25:28

    Christmas … but which year? 🙂
    Any progress with A33, H3 and Mediatek-based OlinuXinos?


    • OLIMEX Ltd
      Oct 16, 2015 @ 18:02:30

      hopefully this year, you are right that we didn’t release new boards for a while and there is good reason for this I may blog later about this 😉 but in short – it become a little bit boring. Each 6 month Allwinner make new chips, but I do not see the point to make yet another board with the newer super duper bla bla processor which basically does same as previous ones and have zero linux support. We still have no mainline support for earlier SoCs with video and 3D acceleration, 99.99% of the use cases for our customers already are met with A20- chip which have everything. What is the point to release board for each new SoC which come? 4 core, 8 core with no Linux support, binary blobs and propertiary uboot. Who need this? I do not see any reason to waste time on this. It was interesting at the beginning – to see if we can make it. Now we know that we can make it, so no challenge, neither fun.
      This is why we spent most of this year working on ESP8266 and things around it. It’s more interesting than to make yet another board with known core and few more cores than previous one, running three years old Kernel 3.4 and with nobody interested or capable to move all drivers to more decent kernel. Allwinner obviously is not interested to support and pay for mainline kernel, their focus is Android which bring them back money. They joined Linaro but what they do is to move H8 and A64 support to Kernel 3.10 to may release Android 5 nothing else. Mainlining is obviously too much work someone from linux-sunxi to do it in the free time, so it’s doomed circle. 🙂


      • Luca
        Oct 16, 2015 @ 21:58:50

        Then you could nove away from allwinner and try the new Intel atoms which are surely more Linux friendly and they don’t cost too much.

      • OLIMEX Ltd
        Oct 16, 2015 @ 22:15:30

        what is not too much? $5? or $50?

      • Bobby
        Oct 17, 2015 @ 12:52:43

        I bought A20-OLinuXino-MICRO more than 2 years ago and it is still the best overall and fastest OLinuXIno today. Maybe you should really focus on IoT and create only 1-2 new OLinuXino’s each year because what we have now as announcements from Olimex are:
        A31-OLinuXino – development began in 2013 but later was officially cancelled.
        A80-OLinuXino – announced in summer 2014, nothing since then, probably is dead too
        A33-OLinuXino, H3-OLinuXIno, A64-OLinuXino – recently announced, some of them are probably going to be cancelled too, low probability something to be released in 2015.
        And you are absolutely right that Allwinner release bazillion chips every year and the difference between them is so small. Even in their first 64 bit chip they still Mali 400 GPU. Come on, it is 2015! They used the same in A10 from 2011! And it seems the open source LIMA driver won’t be finished anytime soon.

  2. zoobab
    Oct 16, 2015 @ 17:53:47

    I have an A10 cheap laptop, the worst keyboard and screen I have ever seen. Please make an effort by choosing good components for the screen and the keyboard (I know a lot of geeks who love the IBM/Lenovo laptops for their keyboards).


  3. novalkar
    Oct 16, 2015 @ 20:53:43

    http://pdadb.net/img/psion_s5mx_pro.jpg << that body would be awesome 😀


  4. Morgaine
    Oct 17, 2015 @ 00:25:44

    Unless you spend a LOT of money on quality materials, your laptop’s hinges and flexible power/signal connectors will die quite rapidly, maybe not even lasting the 1-year default EU warranty period. Returns will cost you money and pain, and your customers won’t be all that thrilled either. This happens even to quality laptops, and you won’t be able to afford quality materials given your price targets and low volumes.

    So I suggest an alternative. You’re very good at LiPo circuitry, so why not keep the keyboard fully detached and LiPo powered, and make the main board and display effectively a tablet. The two could talk to each other using your ESP8266 modules, or the old and now very cheap Bluetooth. They wouldn’t even need to dock physically, saving you money on the connector as well.

    No hinges, no flexing power and comms connectors just dying to break, and we’d end up with one of those very convenient tablet / laptop convertibles instead of a plain and more cumbersome laptop.



    • LinuxUser
      Oct 18, 2015 @ 09:37:16

      On other hand,
      1) Radio things are still consume some power.
      2) It makes system design more complicated and more costly.
      3) You still have to supply power somehow… and either you have to put some wires or separate battery, with all inherent disadvantages of tracking TWO battery charges and sudden deaths of keyboards.
      4) Wireless much less reliable. Wi-fi can suffer of interference and get really laggy/packet lossy.
      5) Wi-Fi haves some dumb protocol-level shortcomings. Its easy to use software like MDK3 to kill ANY wi-fi activity in the area, using cheap and common equipment (its not jamming, its smarter protocol-level attack, which needs far less TX power). Anyone can do it at will. Don’t you mind you will completely lose control of your device, should interference or unfriendly aerial behavior happen?
      6) Throwing passwords into the air could be not best idea ever either. WEP and TKIP already shown us some examples. And if one manages to take over this wireless link, its also possible to completely take over whole device. Since it looks rather attractive in terms of “free lunch” … ask google for “zombie drone” to get idea what would follow.
      7) Every time you press a key, you’ll throw at least 2 MAC addresses to the air. Nice privacy killer device 🙂

      And from more realistic point of view, wireless mices and somesuch are usually using NRF2xxx low-power RF transmitters. After all, full blown wi-fi is kinda overkill for such application.


      • Morgaine
        Oct 18, 2015 @ 17:13:40

        LinuxUser writes:
        > 1) Radio things are still consume some power.

        Very little in the current state of the art. My latest RF mice last 18 months on a primary cell, and even my many-years-old Logitech mice and desksets last a month on a recharge.

        > 2) It makes system design more complicated and more costly.

        Yes, but not by a lot. Olimex makes many kinds of low-power RF modules so they have experience in keeping the power consumption down. What’s more, they buy the RF devices with volume discounts already. You can’t compare the RF costs against zero anyway, because the flexible connectors required to make a laptop hinge last a few years would need to be high quality, so that has a significant cost too.

        > 3) You still have to supply power somehow.

        I addressed that in the first post. Olimex makes great LiPo-powered subsystems, and they add them to even the cheapest boards so the cost can’t be high. It wouldn’t need charging from the screen unit — everybody has a USB charger already, often a dozen of them. 🙂

        > 4) Wireless much less reliable. Wi-fi can suffer of interference and get really laggy/packet lossy.

        The keyboard and screen unit are right next to each other, so such issues are rarely severe unless you live next door to a radar station. Also, throughput doesn’t matter in a HID-to-screen RF connection, even retransmits have a generally imperceptible effect on HID latency. The fact that everybody uses RF HID gear these days and yet complaints are quite rare shows that RF issues are not a big problem in this low-speed application.

        > 5) Wi-Fi haves some dumb protocol-level shortcomings.

        It doesn’t have to use wifi, there are various RF alternatives. Non-BLE Bluetooth devices are probably even cheaper than an ESP8266.

        > 6) Throwing passwords into the air could be not best idea ever either.

        All the RF technologies offer reasonable encryption these days, and this is a very short-range application anyway. Same answer for point 7).

        It’s worth pointing out that almost every one of your objections is moot anyway since the SoC has no built-in Ethernet and the mainboard will most often be connected externally by wifi anyway.


      • LinuxUser
        Oct 20, 2015 @ 06:47:59


        1> Very little in the current state of the art.
        To some degree its true. And your mouse likely uses NRF2xxx (Nordic). Most logitechs are nothing more than mouse sensor -> SPI -> uC -> NRF2xxx low-power transceiver (probably on SPI bus as well). But I bet Logitech spent a lot of time on proper power management, doing DC-DC right and so on. And even then Logitech mouse could sometimes annoy you by some rare lockups which require power cycle. Or it just being slow to wake up from lowpower mode, taking about 5 seconds of circlings like mad. And if you use mouse a lot, it wouldn’t actually last for many months. These numbers are for quite inactive use.

        2) As for costs, sure, if you make mouse which you can sell at $30+ in millions, change it a bit and repeat the same, it would offset R&D costs. But mentioned devices aren’t THAT popular. Not to mention everyone can just plug their Logitech receiver into USB port and rock-n-roll. Yes, it works under Linux. Though I’m not sure if their receiver is state of art in power management, but it should use Nordic and uC as well, and I guess main system CPU would offset it anyway. And pairing device… uhm I wonder: if I had keyboard and it loses pairing, like Logitech mices sometimes do, what the hell I should do? Go bring some “real” keyboard to fix it? TBH I still have old Logitech receiver and mouse which lost pairing. This receiver talks some old flavour of protocol barely resembling modern unified receivers. It somewhat different/old age. So software fails to work with it. I’m just unable to pair this mouse and receiver under Linux. So I have 2 useless hardware parts.

        3) No, you didn’t. It comes down to just 2 options: either you put wires, or you add yet another battery into system. Which is still separate battery and can lose it charge. It takes efforts to write software to monitor this charge and there is still chance battery would die when you need it most. For that reason I can understand wireless mouse (power cord really hurts mouse movement). But perfectly fine with wired keyboards, and I consider the fact they do not need attention as major advantage.

        4) As for interference… I’ve seen at least several cases when Logitech mices were jammed by somethins and barely responding. Even when receiver put 10cm from mouse. I guess there was some strong interference. That’s where you curse the fact your mouse is wireless. Even less fun if keyboard goes down like this.

        5 & 6) Good luck separating marketing bullshit from real security. You see, I pwned about dozen of wi-fi routers to show my friends how “secure” their setups are. Without even touching them. I have reasons to expect other protocols to have more or less the same “security”. You see, its easy to yell in marketing crap about security. But hard to implement crypto right. Then, it runs in some firmware. Which usually lacks source. So nobody could even check how badly it implemented at all. Then, you see, yesterday SHA-1 has been okay. Today it nearly completely broken already. I guess soon we’ll see a dozens of fake SSL certs :). I wouldn’t bet on such “security”. Basically, encryption is hard to get it right. And even fully open designs are subject for review and replacement. It is quite troublesome when its embedded device with firmware. And all these ESPs and bluetooth are looking pretty much like minefield.

        As for ethernet… well, there is USB, it can be everything. Right now I’m actually trying to get my allwinner board to pretend its “Ethernet adapter” to get extra networking link to my PC, usb-otg is kinda cool :). I really like idea when one wire can both supply power to board and provide connectivity. One can also connect usb-to-ethernet adapter to usb host to get “real” ethernet. That’s what Raspberry Pi does, btw. Its SoC can’t do Ethernet. So they put USB hub IC on board and it also does some 100Mbps Ethernet. It worse than native Ethernet, but better than nothing.

      • Morgaine
        Oct 21, 2015 @ 16:28:36

        Re 1,2,4,5,6): I don’t know why you’ve had such bad experiences in all those areas, but it doesn’t mirror my experiences at all — pretty much everything works here, and well. Perhaps it’s because I use only Linux and compile virtually everything from source, which tends to ensure that everything hangs together well and is a known quantity. And maybe your local RF environment is harsher than mine. Regarding security, there is no panacea. As Schneier has often said, it’s only as good as your local security practices and inversely proportional to your convenience.

        Re 3): Indeed, the detached keyboard would have its own power supply. That’s why I mentioned Olimex’s good use of LiPo in multiple products, and also mentioned that everyone already has a USB charger (and USB leads) so they wouldn’t need to be supplied. That keeps costs down.

        Incidentally, a detached keyboard for the display device could be an existing COTS product, such as https://www.olimex.com/Products/OLinuXino/Accessories/KBD-BT1/ . There is no need for everything to be reinvented just for “looks”. We’re not Apple’s target audience.


  5. Freire
    Oct 17, 2015 @ 21:19:09

    A64-OLinuXino(OSHW) + Remix OS(Free*), Perfect!

    Remix Mini PC also uses the processor 1.2GHz quad-core cortex A53 (64-bit) Allwinner.




  6. LinuxUser
    Oct 18, 2015 @ 09:07:48

    Hmm, does it really lacks even Ethernet? Dammit, its cheap, and powerful. But what one have to do with 4 superb cores, if they nearly lack any IO?

    And what about version without wi-fi/BT/etc? It increases price but I have some use cases where I do not really need wireless or even explicitly do not want it due to risk of remote hacks.


  7. Trackback: Open Source Hardware-based AArch64 board from Olimex | Firmware Security
  8. SK
    Oct 19, 2015 @ 01:52:24

    Why don’t you wait for H64 which would have native Ethernet?


  9. Luca
    Oct 19, 2015 @ 13:13:18

    In response of a your response (the REPLY button is missing) regarding Intel atom SOCs:

    > what is not too much? $5? or $50?

    If you see the new Atom X3, it costs $20. Okay, this is more than the Allwinner but the Linux support is better hands off.


  10. ssvb
    Oct 20, 2015 @ 15:45:46

    The absence of ethernet is bad for a development board. Having only a single USB host connector is also bad. Would it make sense to add more USB connectors and also ethernet via an additional chip, just like some of the other development boards do? How much would it increase the price?


    • OLIMEX Ltd
      Oct 20, 2015 @ 17:08:20

      for A64 we will make just WiFi + BT, when H64 is released it will have native Ethernet and will have good performance than going through USB


      • SK
        Oct 20, 2015 @ 21:30:30

        A onboard USB hub, so we can have > 1 USB host would be nice, though.

      • Palo
        Oct 21, 2015 @ 12:07:24

        USB2.0 or USB3.0?

      • ssvb
        Oct 22, 2015 @ 18:43:46

        True, but as you mentioned it yourself, “No indications that H64 will be ready any time soon”. Raspberry Pi and ODROID boards solve this problem by using LAN9514. Native ethernet is surely faster and cheaper when it is available. But IMHO having no ethernet at all makes the board less attractive when compared to the competitors.

        And now it’s a race between companies, competing to see who brings a reasonably priced 64-bit ARM board to the market first 🙂

  11. vonfritz
    Oct 21, 2015 @ 09:52:59

    Why not drop Allwinner and go for something more linux friendly?
    Actions semi S500 and S900 seems seems more linux friendly.
    Sources are out before Board is available. 🙂
    And H64 seems to be the next orangepi plus4 board.


    • OLIMEX Ltd
      Oct 21, 2015 @ 10:11:51

      the answer is in your question -> “seems” more linux friendly
      just read the Odroid forum about the problems with the video


    • ssvb
      Oct 21, 2015 @ 16:59:23

      Contrary to some recent hate instilling propaganda articles, Allwinner is still one of the most Linux-friendly ARM SoC vendors around. They are providing documentation and source code for everything that is essential.


  12. Matej
    Oct 22, 2015 @ 02:26:41

    When there is no Ethernet, SATA but USB host. Can you add 4x USB port? People will buy USB-Ethernet and USB-SATA adapters… +2x for keyboard and mouse or two gamepads… 2xdouble usb and 2xdouble usb?
    Can PCB mask be white and white color LED? As Christmas are comming?
    Haha it can have name Olinuxino-Yeti-A64 or something like that.


  13. OLIMEX Ltd
    Nov 03, 2015 @ 16:06:37

    I guess not, as this laptop looks like plastic toy, I think for something more stylish and reliable


  14. liestdochkeiner (@liestdochkeiner)
    Nov 03, 2015 @ 16:08:02

    Concerning a laptop: it should be very easy to use an A64-OLinuXino in a Pi-top (see pi-top.com), right?


  15. 99guspuppet
    Nov 06, 2015 @ 00:07:06

    Any chance of adding a USB3.0 port ?


  16. ekimiablog
    Nov 06, 2015 @ 00:09:47

    Open source + Mali ? How?


    • OLIMEX Ltd
      Nov 06, 2015 @ 08:00:44

      enlight me please how Mali is related to OSHW?


      • ekimiablog
        Nov 13, 2015 @ 10:43:14

        You’re right. But OSHW with Proprietary graphic drivers is a waste of time

      • OLIMEX Ltd
        Nov 13, 2015 @ 10:54:14

        every one is free to have his own opinion 🙂
        for some one OSHW without knowing the electron orbits inside the SOC may be also waste of time but I personally do not care about this concern.
        OSHW main idea is to share knowledge and people to may learn from your work and give you feed back so you can learn from others too.
        That there is *some* software which hides some stuff is not problem for me, if you do not want to have binary blobs just do not run such software, nobody is forcing you to do this. Use software encoding or do not use graphics at all if you are so radical about what software is run on given machine.
        I do not understand HOW this relates to the Open Hardware ideas and how this makes useless the knowledge how one device is designed as hardware and the share of this knowledge with the others.

      • ssvb
        Nov 13, 2015 @ 10:52:48

        But you don’t need to use proprietary graphics drivers. Mali is only a 3D engine and you don’t strictly need it unless you want to play 3D games or have fancy desktop compositing effects. Basic 2D desktop works fine without Mali.

        Moreover, there has been a significant progress achieved reverse engineering Mali and you can already run simple 3D demos (such as a spinning cube) without any proprietary blobs. Implementing a feature complete Mesa driver is a significant part of work, which is yet to be done. Along with implementing a rewritten kernel driver, which good enough to be accepted in the mainline kernel.

  17. Slackstick
    Nov 10, 2015 @ 18:57:57

    Attach 4GB of RAM and 32GB EMMC and you have my order.


  18. Jakša Tomović
    Mar 21, 2016 @ 11:29:42

    hello where can i buy A64…is it in sale allready, if not when can i expect it to be? Thx cant wait to buy one 🙂


  19. Jakša Tomović
    Mar 21, 2016 @ 11:31:29

    Hi…can you tell me can i buy it allready and where, if not when can i expect it ? Thx and i am looking forward


  20. Maroci
    Jan 08, 2021 @ 13:40:55

    Hi, I assembled the Olimex Teres I laptop. Iz followed the instructions and everything workded fine at first attempt.
    however i have an issue either with the teres pcb1 a64 or the screen.
    I was trying to expand the memory of /boot with gpart and I feel I made a mistake overwriting a part of the memory label as unknown. When i boot the laptthe screen it was completely black.

    So I open it again and I found the connector of the screen a bit unplugged.
    I plugged it again but the screen still doesn’t go on when I boot the computer.

    I would like to understand how I can get help to understand if either the screen is damage because the connetor of the screen unplugged or I erased the memory to boot the board.

    Please help me.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: