Is it time for Open Source Hardware x86 OLinuXino?


We have been so focused on ARM that somehow we have missed the Intel latest development.

Yesterday browsing the net I saw $100 tablet with X5-Z8300 and check the datasheet. It looks good!

Quad Core 1.84Ghz 64-bit data bus “Cherry Trial” Atom with support for up to 4GB DDR Ram, USB 3.0 (!) we guess working as on A80 USB3 is one of the broken peripherials. PICe, MIPI camera, display etc.

All this come in handy BGA 17×17 mm 0.65 package not quite different than Allwinner A64 part.

And of course it comes with first class documentation you can see product page and documentation.

Intel write recommended price for tray $20 in their site, but brief check shows that it could be sourced from China at less than $15. Much better deal than the obsolete iMX6 quad core SoC for instance! Still x3 times more expensive than Allwinner A64 but probably x10 times more powerful.


  • first class SOC vendor 14 nm process;
  • well documented;
  • nice features;
  • removes the cumbersome ARM obsolete Android Linux Kernels problems;
  • you can work with the latest vanilla Linux Kernel;
  • we guess Graphics and Video drivers are also mainlined;
  • can run Linux, Windows, Android whatever you want;
  • can be build to work with standard  64bit computer memory modules, so you can put whatever amount of RAM you want up to 4GB;


  • probably more power hungry than ARM, but maybe not, the tablet I was looking at was with 4000mAh battery



59 Comments (+add yours?)

  1. ern0
    Nov 19, 2015 @ 09:57:10

    What is your estimation for the “street” price, compared to similar ARM models?


  2. ARM Solider
    Nov 19, 2015 @ 10:07:24

    I hate Win an INTEL. Forever ARM and Linux…


  3. Mossroy
    Nov 19, 2015 @ 10:28:06

    Mainline kernel support would be really great.
    But IMHO it’s important to keep low heat, so that it can still be fanless


  4. onion
    Nov 19, 2015 @ 10:33:02

    What about something MIPS based ? Loongson perhaps?


  5. jlucius
    Nov 19, 2015 @ 11:10:41

    + good documentation
    + mainline support
    + no mali / powervr binary blob issues

    You might need to get into EFI boot stuff, but I think there are intel sdks for that. You can get an Intel developer account for free and have a look at datasheets etc.


  6. progmetalbg
    Nov 19, 2015 @ 11:20:27

    I think you need UEFI or BIOS chip from 3rd party vendor in order to boot the system which is an additional expense. I might be wrong though …


  7. Allen
    Nov 19, 2015 @ 11:40:42

    Two or more ethernet port for use with pfsense and asterisk distro. 🙂 🙂


  8. cnxsoft
    Nov 19, 2015 @ 11:45:04

    If you could also checkout x7-z8700, which should not be that much more expensive. It depends how flexible you want your design to be. See,85475 for comparison

    About power consumption, I checked Intel Atom Z3735F before, which should have the same TDP as x5-Z8300, and it had only slightly higher power consumption compared to Rockchip RK3288 and Amlogic S812 ->


    • LinuxUser
      Nov 22, 2015 @ 00:25:17

      But, somehow, I happened to had x86 tablet. And then I ditched it in favor of MTK based one… because it lasts like 2x longer in active mode when I using it, and like 10x longer when it in “sleep”, while retaining network connections. What is most ironic: SAME BATTERY SIZE.

      Basically, Intel’s power management is an utter crap. It is some overcomplicated clusterfucked mess inherited from DOS ages and then they attempted to amend it a lot, to make more resonable. At the end of day, it causes nasty bugs and glitches here and there. Yet, it happens to be very bad at actually saving power. Basically x86 can’t be low power and remain operational (e.g. keeping a network connection). It could be less issue in embedded designs (where I would rather be unhappy about uefi), but crappy power management is really bad in tablets, smarphones and somesuch. Whatever intel mumbles, but crappy & cheap MTK haves order of magnitude better power management. So I had to return x86 tablet: its no fun when device charges longer than discharges, dammit.


  9. Bib
    Nov 19, 2015 @ 12:12:59

    Ok for the Pro but you should add for Con :
    1- No longevity program with Intel Atom,
    2- BIOS (UEFI) not GPL’d and consequently the ability to distribute hardware.could be jeopardize.


  10. LinAdmin
    Nov 19, 2015 @ 12:40:16

    I have tried for a long time to get a small file server based on ARM, but have not yet found a board with GHz LAN and 1-2 SATA ports which does read and write via LAN at about 100MB/s and power consumption of 12V 2A included WD red disk running Jessie and price below 100$.

    Each and every board in the past showed some important deficiency, and if you can find such a solution I certainly am not the only one interested in it!


  11. zoobab
    Nov 19, 2015 @ 12:48:06

    Let’s look at the complexity of booting an OS under an X86 architecture, compared to an ARM one. There is 30 years of backward compatibility legacy in the current X86 BIOSes, while uboot simply jumps to an address and done.

    But if you persist, please have a look at using a free bootloader, such as coreboot (if it has support for that chip).


  12. Petr Moses
    Nov 19, 2015 @ 13:06:28

    Street price of CPU should be about 5 USD. Question is, how low could you go with price, if you can beat big brands on price and support.
    You can invent new board format and sell notebooks with replaceable mainboard, with processor up to 10W, but economics is questionable. Up to now you sell to hardware makers, consumer market is different league.


  13. stratic
    Nov 19, 2015 @ 13:17:53

    x86 OLinuXino will be great for me. All current OLinuXino boards lack power for my applications.

    “Cherry Trial”: is it the next generation SoCs from Intel ? 🙂


  14. Pau Garcia i Quiles
    Nov 19, 2015 @ 13:35:47

    Linux support is not so great yet.

    Kernel 4.3 includes some support but there is still serious trouble with the audio (no ALSA support yet, no matter what you do).

    With the existing Bay Trail and Cherry Trail devices, there is also the problem of the wifi+BT device. Chinese manufacturers tend to use either Realtek or Ampak (Broadcom) chips which are not supported yet. There is out-of-kernel support for many of them but firmware needs to be provided, the NVRAM configuration info needs to be loaded (the brcmfmac driver does not support loading the NVRAM from the UEFI, although I have a patch that should make that work), etc

    Also, there is the problem of the UEFI. You want to run a 64-bit operating system but usually the UEFI on these systems is 32-bit for several reasons. Some of them are explained here:

    That means none of the existing Linux distributions even boot on Bay Trail/Cherry Trail systems. I have a customized ISO of Ubuntu 15.10 where all those problems are sorted out (minus the audio, which as I said is not solved yet by ALSA). Recent snapshots of Debian with the latest debian-installer also support booting on mixed environments with 32-bit UEFI + 64-bit CPU.

    It would be great if the OlinuXino x86 would boot from coreboot (or even u-boot). It’s already supported and probably less cumbersome to implement than UEFI:


  15. dmitriybeykun
    Nov 19, 2015 @ 13:43:06

    Hello Tsvetan,

    I’ve successfully ran ArchLinux + KDE 5 Plasma on my Acer 8W $100 tablet

    8W is Z3735G tablet on Baytrail platform, running 1GB of RAM and 32GB eMMC storage. Touchscreen is Goodix, LCD is connected via MIPI-DSI. UEFI is 32 bit.

    Problem is, mainline linux support is insanely bad and there are no alternatives to it, except Yocto which seems to have paid support or something like that. I could not find proper official Baytrail BSP for Linux.

    List of problems:
    1) UEFI is 32bit and this is bad due to most linux distros not having grub compiled in 32bit mode. Solved this with manually compiling grub into bootia32.efi and installing it afterwards. Also done same for rEFInd.
    2) Backlight control is not working, it should be and i915 have Valleyview support, but it seems that something is missing, probably ACPI related
    3) ACPI battery status is not working due to missing support for ACPI BMOP region. BMOP region is tied to X-Powers AXP288 PMIC (yay, same as on Allwinner boards), but working directly with it is not the way it should be done on x86. Intel submitted initial support, but then decided to drop the AXP related bugs like this one This ultimatevily shows Intel’s attention to mainline on Baytrail. They even done SharksCove board but mainline is still missing, while nothing prevents it to be done
    4) ACPI buttons do not work, probably can be worked around with DSDT changes or custom drivers same as for some notebooks
    5) UEFI is insanely bad, it is done by INSYDE and can store BootRecord only matching path /efi/boot/bootmgr.efi or something like that, any other pathes will be dropped from BootRecord list, though, BootNext actually works. Bravo, INSYDE.
    6) Sound doesnt work even having all the needed drivers for realtek chipset and binary firmware blobs for the Intel part
    7) HPET is disabled on Baytrail because it is unstable.
    8) Obviously, suspend/hibernate doesnt work
    9) Goodix touchscreen registers cursor movement, but does not register clicks, might be solveable with proper X.Org setup
    10) atomisp MIPI-CSI driver is not available in mainline and works only with couple cameras

    So now I want to ask you, are you ready to invest 3-5 engineers into this fulltime for atleast 2-3 months before they polish out all these problems? Without any guarantee that problems are solveable? I’m not even going to start about the technical documentation for Intel’s SoCs and debug process.



    • jlucius
      Nov 19, 2015 @ 16:23:04

      1) This is a vendor decision, so Olimex could support 32- as well as 64 bit.
      2)-5) ACPI on most cheap tablets is fu ..nny to use, most vendors have errors in that part causing much trouble on linux.
      6) Intel Sound driver is mainline (with binary firmware) and is working. One some tablets it is connected to another port to what the driver excepts and needs to be patched.
      9) Touch should work fine, click works without problem here.

      To use Yocto you need the meta-intel layer and the valleyisland-64 machine. If you need some starting points you can ask me, but thats a self-build distro. Ubuntu should run on a baytrail tablet with some restrictions or use the version of linuxium.

      I was also hoping that baytrail is supported better in mainline, but most stuff works already (better than A20 mainline)


      • dmitriybeykun
        Nov 19, 2015 @ 16:42:54

        > 1) This is a vendor decision, so Olimex could support 32- as well as 64 bit.
        any links to the docs for this?

        > 6) Intel Sound driver is mainline (with binary firmware) and is working. One some tablets it is connected to another port to what the driver excepts and needs to be patched.

        this needs to be handeled automagically by the driver itself. Not speaking about that couple months ago sound module did not work at all if you did not plug in HDMI cable, code was done that way that if it fails to init HDMI sound it completely bails out 🙂

        > 9) Touch should work fine, click works without problem here.
        Probably Goodix specific problem.

        > Ubuntu should run on a baytrail tablet with some restrictions or use the version of linuxium.

        I checked all ubuntu versions and they have same problem: mipi-dsi clock output has jitter. This is solved by using kernel that is close to mainline or drm-next.

        > I was also hoping that baytrail is supported better in mainline, but most stuff works already (better than A20 mainline)
        Better, but not enough good to use it daily. I would live up with all the sound/touch/ACPI crap if voltage scaling/battery and charging would work…

      • jlucius
        Nov 19, 2015 @ 18:29:36

        Depends where the firmware is from. My understanding is that the hardware vendor does provide the firmware and can decide to compile it 32-Bit or 64-Bit. In any article regarding this topic it´s always the term “hardware vendor” and the hardware vendor does need to provide the correct tables for ACPI and hardware setup, so it should be correct? (there are very limited information available about that topic, so thats actually my guess, any information about this would be helpful).

        Sound: I think there was actually a RFC how to detect automatically the correct port, but noone picked it up. Seems the development train already moved to the newest intel platform. So far I can live with patching my kernel.

        I also have a Goodix Touchscreen and no trouble with mainline kernel. More of an X / Calibration Problem?

        There are many problems with the baytrail platform and linux, but having a nice devboard from olimex probably attracts more developers and we are talking about the next gen cherrytrail here.

  16. Luca
    Nov 19, 2015 @ 13:57:54

    I have a Minix Z64A, based on an Atom Z3735f. I’m running a custom Ubuntu 14.04.3 made by the famous Linuxium, so I am using the older Kernel 3.16 but with Audio+Wifi+3D working perfectly.

    I’m very happy, I bought the minix used for only 80 EUROS and it consumes only 4watt TDP!


  17. anonima
    Nov 19, 2015 @ 14:47:54

    Intel and Linux are good friends. BIOS.


  18. Morgaine
    Nov 19, 2015 @ 14:49:58

    Wow, what an excellent set of responses, highly informative. Unfortunately they summarize quite unanimously as “It would be a research project of unknown duration and success not guaranteed.”

    It seems that ARM doesn’t have much to worry about for a few years yet.


    • jlucius
      Nov 19, 2015 @ 16:15:50

      But these are mostly Linux related. Mainline for Allwinner A20 is not completly done yet: battery support is still missing, graphics driver is not done (not speaking about hardware acceleration at all), flash memory support has just hit mainline, not supporting all functions yet. Basic support for H3 is just being done at the moment. Most other ARM devices are stuck at old kernel versions 3.4, 3.12 or at most 3.14, whatever the vendor supports, so Intel is not that bad comparing to that. Android and Windows should run fine.


  19. SK
    Nov 19, 2015 @ 15:02:11

    The specs say this processor can have max 2GB, and the base clock is 1.44GHz


  20. xxiao
    Nov 19, 2015 @ 16:23:19

    No Ethernet? This is the biggest drawback. Also no SATA? Can it be fanless? Does it have industrial class version? I for one do not like Intel chips at all but its ATOM series do have higher CPU processing power than most ARM/MIPS which is important sometimes.


  21. jlucius
    Nov 19, 2015 @ 16:35:04


  22. Kenny
    Nov 19, 2015 @ 16:53:46

    Can you comment on the closed-ness of it? Will it have SMM and/or ME built in? What about microcode blobs?

    Personally I’d be much more excited to hear you were looking at


    • Filipe
      Nov 19, 2015 @ 22:07:52

      Given Intel’s Management Engine, TST, vPro technology, etc., I don’t think they have a good track record. ARM isn’t perfect either, but seems more willing to cooperate?


      • Matyas
        Nov 20, 2015 @ 13:40:18

        Yeah, that’s what I brought up on Twitter. Intel is not particularly known for its openness.

      • LinuxUser
        Nov 22, 2015 @ 00:14:47

        They also always trying to supply proprietary UEFI. Which is huge, bugged, inclined on PC-like usage and it is nearly impossible to ditch it in favor of more sane and customizable boot loader. Overall, Intel designs are complicate, backdoored and inflexible. They barely managed to made it to tablets by paying (!!!) chinese vendors so they start using their ICs at all.

    • k.
      Nov 21, 2015 @ 12:56:05

      Such projects as lowRISC really deserve support. Intel does not.


    • LinuxUser
      Nov 22, 2015 @ 00:37:01

      Same here! Backdoors are no-go for serious and resposnible engineering.


  23. Felipe Balbi
    Nov 19, 2015 @ 17:57:54

    Go for it, just, please, make USB3 peripheral available 🙂 They use the driver I wrote (drivers/usb/dwc3/) and I have been longing for an x86 platform to test. Currently only ARMs around.


  24. jonsmirl
    Nov 20, 2015 @ 03:37:31

    How about an Olinuxino based on a cell phone chip with at least a 3G SIM card modem? Like a Mediatek chip.

    Or maybe this Intel SoFIA one? Whole 3G phone is $34.

    2G is being shut down in US so I wouldn’t bother with it. LTE is too expensive currently.


    • LinuxUser
      Nov 22, 2015 @ 00:09:57

      Hmm, if just SoC costs whopping $34… I guess at the end of day, $5 allwinner SoC + $15 HSPA modem Olimex sells on their web site is quite a deal :). So “just $34” could be kinda relative term. Don’t you mind most ARM boards are like $35 or below these days? And boards above $50 are getting rather unpopular these days due to emerging competition.


      • jlucius
        Nov 22, 2015 @ 20:50:36

        Nope the whole phone costs $34 including boad, processor, ram, flash, display and case.

      • jonsmirl
        Nov 22, 2015 @ 20:55:16

        Linuxuser, missed your reply earlier — that is an entire phone x86 CPU, 512MB RAM, 4GB flash, 3G modem, 2.4Ghz wifi, BT 4.0, 800×480 touch LCD, battery, case, Android 5.1 – all for $34.

  25. FergusL
    Nov 20, 2015 @ 11:25:56

    To answer your question: YES, it is time for it, definitely.
    It’s already being done actually:
    For some specific areas like realtime audio processing, ARM is a nightmare, you have to compile almost everything because of lack of precompiled packages.

    + we trust you you’re not going to follow the utterly stupid Raspberry Pi form factor and 40 pin header layout.


    • LinuxUser
      Nov 22, 2015 @ 00:02:00

      …and why do you need to do “realtime audio processing” on such a strange hardware? Not to mention it is rather specific task which could want a crapload of system tuning anyway.

      And btw, ARM debian repo got virtually all packages you can have on x86. With maybe very few omissions.

      At most I can imagine building low-latency kernel. And when I build kernel myself, I can trade speed (and idle power consumption) vs latency, hitting the point I want. E.g. lowest latency assumes more CPU wakeups and more power consumption, as well as womewhat reduced performance. OTOH server tuning implies bad latency but good bulk performance. Actually there’re quite many possible combinations even of most obvious features like HZ (task switching timer rate) and ability to preempt kernel. Combination of these two gives like a 12 possible ways to build a kernel, or even more. Then there is CPU frequency switching, and I guess ARMs are orders of magnitude more predictable: just lock frequency to single desired value and rock-n-roll. OTOH, x86 comes with messy clusterfuck of ACPI calls and power states, implying quite high latencies between states and low predictability.


  26. SK
    Nov 20, 2015 @ 15:03:19

    There is a newer model announced for Q1-2016, with slightly higher boost clock frequency (would help in single threaded usage) – x5-Z8350


  27. FergusL
    Nov 21, 2015 @ 18:57:46

    I assume they should be, it’s really just a “product change” :

    To extend on the ARM or Intel debate, I think it is a healthy thing for the market that Intel gets in these low-power low-cost areas to reduce the ARM monopoly. It is as healthy as seeing ARM getting up the other way towards the laptop/desktop and even server segment with the same idea of lowering the impact of another IP provider monopoly.

    I believe that us, users and manufacturers of open source SBCs must realise our role is also to balance our choices regarding these two major brands.


    • jonsmirl
      Nov 21, 2015 @ 19:22:57

      For a long time Intel has not wanted to compete against ARM in the low end. That’s because making $5 x86 processors would erode their brand image and put a lot of pricing pressure on those $2,000 CPUs that Intel sells with 90% margins. I think they are only doing these cheaper x86 chips to counter the ARM64 threat. ARM64 has a lot of potential to do damage in the high end, high margin market.


  28. LinuxUser
    Nov 21, 2015 @ 23:50:37

    Speaking for myself, I would expect
    * Overcomplicated system design.
    * Awful prices.
    * Proprietary UEFI shit.
    * Very hard to make it doing what I need rather than what Intel needs. E.g. UEFI can do something unwanted on some interface. And there is no need to get rid of it.

    So I wouldn’t mess with these for sure and would rather wait for more cheap and neat ARM boards.


  29. Jakša Tomović
    Apr 06, 2016 @ 22:42:45

    So will you make any board with Intel Atom? I think that would be awesome!!!


    • OLIMEX Ltd
      Apr 07, 2016 @ 09:32:10

      First I though it’s a good idea, but it seems Intel is not interested in OSHW design as they do not respond to any of my e-mails and local Intel distributor offer price for 1000 pcs Intel processors at 30% above the Intel’s recommended sale price for 1 pce, so it seems Intel do not sell volumes in Europe and focus only in Asia sales, production with these conditions make no sense. One other thing we saw is that Intel Atom Linux support is far from complete, all Chinese laptops and tablets run Windows 10 in 32 bit mode as Intel provide only 32 bit UEFI bootloader 😀 running Linux on Intel Atom is possible but lot of drivers are broken, so for the moment 64 bit ARM is much more promising.


Leave a Reply

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

You are commenting using your 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: