Home > Hardware > Office, Factory, Business Model, and Ambitious Plans of Shenzhen Xunlong Software, Orange Pi Maker

Office, Factory, Business Model, and Ambitious Plans of Shenzhen Xunlong Software, Orange Pi Maker

Parts of the article have been updated after Steven Zhao update at the end.

Since Steven Zhao is the only one contact person at Shenzhen Xunlong Software, the maker of Orange Pi boards, and the company appears to be focusing on hardware development more than on software and documentation, so at one point in time, people were speculating that it could be one person operation πŸ™‚ Last year, Steven told us there were over 10 persons working in the office. But hey, photos, or it’s fake! We now have a definite proof as Renaud Coustellier visited Steven Zhao in his Shenzhen Offices, and published a report on Minimachines (in French). I’ll provide a summary below, but visit Minimachines website, if you want the full story and more pictures.

First, Shenzhen Xunlong rented a floor, or part of it, in one of the many Shenzhen office buildings, and engineers are working in the typical cubicles, you’ll find in most other companies in China. The 200 m² office is now occupied by 50 employees doing hardware design, testing, and some customer support, and another part of the office is used for shipping parcels to customer. Manufacturing takes place in another factory in a cheaper place in inner China.

Click to Enlarge

The company is not only doing development boards, but also working on projects for their customers such as point-of-sales and game consoles. Those are not Xunlong products, and Renaud was not allowed to take pictures.

Xunlong was apparently initially involved in the development of the first Banana Pi board, and then went on to create their own Orange Pi boards. We also learned the typical development cycle for a new board: two months for the first design, and one month extra for testing, before mass production, so it takes around three monthsin total. Each card has to go through hardware and software tests to pass QA.

Xunlong sells around 40,000 boards per month, but the company is ambitious with a target of around 100,000 units per month by the end of 2017, and a new 5000 m² factory is now being built in Dongguan. Another positive point is that the company plans to hire (more) software engineers in 2017, instead relying  only/mostly on external projects such as Armbian. According to the report, beside Orange Pi board samples, the company also gave $24,000 to Armbian community to help them out. Xunlong also launched an initiative to provide free development kit if you have a specific project for it that would be beneficial to both parties (but I could not find a link).

Renaud and Steven (Right)

Then we also learn a bit more how the company can make boards at such a low price, and this get interesting. This was their conversion:

  • Steven: Orange Pi can purchase a good WiFi component for $1, with which boards will have a good WiFi connection
  • Renaud: Hmmm, if the price of the components is $1, then the price will be higher to the end user, and if you tell me the price, then I know your margin…
  • Steven: No, because today, all Orange Pi boards are sold to the BoM price.
  • Renaud – Does that mean the cost of engineers, office rental, equipment for development and design, and so on, is not reflected int othe price of the board?
  • Steven: Correct.
  • Renaud: But how how is that possible?
  • Steven: We receive subsidies from the government.

There you have it. That explains everything. We all pay about 50% of what we ought to be paying thanks to the Chinese taxpayers.

Beside the new factory, Xunlong has other bigs plans, as while they now sell though Taobao in China, and Aliexpress to the rest of the world, they plan to setup their own shop, where we should be able to buy the board directly from them. They intend to expand to IoT boards with 2G to 5G boards, ARM servers, and so on, and over the next 3 years, the plan is to recruit around 500 engineers, and sell ten times more boards than Raspberry Pi foundation. Maybe subsidies will have long been gone if that happens…

[Update from Steven:

Yes, there are some mistake, maybe it is because our translator wrong expressing.

1. About time arrangement for designing a new board: yes, we need 2 month for first design which also include samples production, but usually we take only 40-45days(30working days). Then after tested at customer side, we need around one month (20working days) for mass production. Which mean for a new product we need 3 months if do not take customer’s time calculating in testing.

2.  The new factory size is only 5000㎡ in Dongguan, and yes, we will donate Armbian $24,000 by the end of this year, for now we have only donated them part of the total amount.

3. About our cost and subsidies from government, yes, we sell most of our product at BOM cost, but we also have some models that still have margin profits, like PC Plus, Plus 2. If we keep running behind our expenses, how could government could support us? Government would not support you, if you have no income, they could only support you parts of the cost, like engineering cost, or part or office renting cost. So most of the cost should cover by ourself, we need to figure out how to how to sell more with margin profit. As you see, we do it now.

Finally, yes, we know what our weakness is, we will try to improve it in the following days.]

Thanks to Nobe for the tip.

  1. lordzahl
    July 15th, 2017 at 15:20 | #1

    ..only in China πŸ˜€

  2. Daniel
    July 15th, 2017 at 16:14 | #2

    Wow seems Orange Pi is here to stay. Let’s hope their software promises come true, I quite like their boards.

  3. Varghese
    July 15th, 2017 at 16:18 | #3

    it is really a good hardware , only headache is display ( raspberry will show something on the screen irrespective of resolution setting) which is very normal for a mobile SOC chip. if you have some software background,it is not a big issue … again kids always expect to spoon feed them & they always complain.

  4. theguyuk
  5. July 15th, 2017 at 17:39 | #5

    @theguyuk
    Hmm, now I remember that one. But the guy in minimachines told they had just started since “several weeks”. While technically 2015 is still several weeks ago, maybe it’s a new initiative?

  6. OvCa77
    July 15th, 2017 at 17:43 | #6

    On this picture I see some new Orange Pi boards: http://farm5.staticflickr.com/4256/35919334535_2d794aecd1_o.jpg

  7. JotaMG
    July 15th, 2017 at 17:46 | #7

    ” the company has gave $24,000 to Armbian community ”

    -> what a waste of money !!

  8. July 15th, 2017 at 17:48 | #8

    @OvCa77
    Good find. I see one Orange Pi Home RTD1295DD Board, and one “Advanced CPU module”. If they start doing SoMs, they’ll sell quite a few more boards.

  9. JotaMG
    July 15th, 2017 at 17:49 | #9

    ” We receive subsidies from the government ”

    -> and, of course, customers receive some spy bots hidden in firmware… really!

  10. theguyuk
    July 15th, 2017 at 17:55 | #10

    @cnxsoft I wonder about Nanopi Friendlyelec, are they government paid or is Orange Pi dumping model going to damage Friendlyelec Nanopi?

    The apply for a dev board link under community on Orange Pi site is only I have found, do they have a different apply page on Chinese version of site I wonder?

  11. PM Nordkvist
    July 15th, 2017 at 18:20 | #11

    Wow! Great read. I always wanted to know mΓΆre about these guys.

  12. Wolf
    July 15th, 2017 at 18:36 | #12

    @cnxsoft

    Excelent news. That RTD1295 dev board may be the one I’m looking for (native SATA+GPIO will make it a win for me)
    I hope you’ll get your hands on a sample board soon for a proper review.

  13. 2mike
    July 15th, 2017 at 18:41 | #13

    // if the price of the components is $1, then the price will be higher to the end user///
    stupid mentality πŸ™
    End users are forced to spend a few dollars to purchase a separate wifi dongle πŸ™
    //We receive subsidies from the government.///
    Now it’s clear, why they do not fix bugs in their boards.
    The only advantage of their boards is the low price and low postage of aliexpress

  14. nobe
    July 15th, 2017 at 18:58 | #14

    @OvCa77 you’ve got good eyes πŸ˜‰
    in another picture, on a desk, we can also see a bunch of very small boards with 2 ethernet connectors side by side. i’m not sure it’s a known product.

  15. Fossxplorer
    July 15th, 2017 at 19:05 | #15

    @JotaMG
    Why do you state that? For me it seems Armbian community puts a lot of effort into supporting different boards.
    Also, $24.000 is nothing wrt. to salaries of IT personal in north Europe in general.

  16. JotaMG
    July 15th, 2017 at 20:41 | #16

    @Fossxplorer
    because even with 1% of that resources, the students at my local university can do a much better job than Armbian does !
    (for instance they already had design a much better and friendly install)

  17. Drone
    July 15th, 2017 at 21:20 | #17

    “We receive subsidies from the government.”

    Enjoy those impossibly low prices while you can you dumb Capitalist Sheeple. Once the Chinese put all your domestic factories out of business, they’re gonna bust you like a shotgun and shove a broken baseball bat straight up where the sun don’t shine!

  18. theguyuk
    July 15th, 2017 at 22:01 | #18

    @Drone
    Which countries factories you thinking of?

  19. lvrp16
    July 15th, 2017 at 22:08 | #19

    Some information here are not necessarily correct. They only have a small subsidy from the government that is open to all tech companies that want to fill out a pile paperwork. The boards are also not sold at BOM.

    It is a kind of misinterpretation of China of sorts thats to be expected. Also, Raspberry Pi is only a small player in the sbc market. The big ones do not make the headlines.

  20. Renaud Coustellier
    July 15th, 2017 at 23:17 | #20

    Thank you to publish my article wand my photos without my authorization.
    I will see what to do, but i am speechless. Incredible.

  21. LinAdmin
    July 16th, 2017 at 02:12 | #21

    I am not in a position to believe all what this french guy said. Maybe Igor will comment?

  22. itchy n scratchy
    July 16th, 2017 at 04:49 | #22

    Missing @tkaisers rant πŸ˜‰

  23. TC
    July 16th, 2017 at 05:43 | #23

    itchy n scratchy :
    Missing @tkaisers rant

    he is probably still typing on it πŸ˜‰

  24. Fossxplorer
    July 16th, 2017 at 08:20 | #24

    @TC
    More likely on vacation πŸ™‚

  25. crashoverride
    July 16th, 2017 at 11:59 | #25

    After receiving $24,000, what is there to rant about? I would say kind things for half that! πŸ™‚

  26. July 16th, 2017 at 13:41 | #26

    > Maybe Igor will comment?

    I can only confirm that donation was promised and that man on the picture is Steven Zhao. Donation is not on the level to help expansions since we already have few full timers (with a real paycheck), from Europe, to keep the project running + there are other fixed costs. Armbian project is non profit and we are probably exaggerated 99% away from making one. In such project there is no need for hiring slave labour @JotaMG to do our “lousy job” better.

  27. July 16th, 2017 at 13:56 | #27

    @Igor
    I manage to make a mistake in such a short notice … “with” a real paycheck is “without” a real paycheck πŸ™‚

  28. theguyuk
    July 16th, 2017 at 15:33 | #28

    @Igor
    For those of use not in your trade can you give us a idea of what costs the project has to meet?

  29. JotaMG
    July 16th, 2017 at 18:51 | #29

    Well Igor…
    I follow your work since early days of cubieboards, and respect you very much for what you have achieved.
    Armbian has done some very positive contributions to the usability of the boards.
    But my opinion is that now Armbian seems to have lost its Karma….

    Regarding you unfortunate and even stupid comment about β€œhiring slave labour”, I’m just a teacher, they are not even students from my classes. Their code is still alpha quality (but works) and they are still undecided what to do with it, but its up to them to decide.

  30. Jon Smirl
    July 16th, 2017 at 18:53 | #30

    @theguyuk

    Fully loaded software engineer in the US capable of working on the Linux kernel costs a US company at least $250K/yr probably more like $350K in SF. Fully loaded includes office, equipment, expenses, all taxes, medical, SS tax, etc. –
    this is not salary. Around $80,000 of that ends up in property, state and federal taxes. Even if you rent you are paying property taxes. Of course all of this varies hugely between SF and Alabama.

  31. July 16th, 2017 at 20:37 | #31

    @theguyuk

    Open source projects can and do create value and debt. Our is no exception in this. Some are repaid with micro donations but only if you manage to make users happy and if they can and want to give something in return. This is extremely hard and rare. And on top of this, we need to fight predators, which abusing even more of our precious time.

    If we don’t count time working on a project – research & development, planning, giving free 1st class end user support, if we neglect this is actually a daily job (for few people) and if we don’t add past investments … we don’t have many costs πŸ™‚ Infrastructure costs are negligible from this perspective.

    Our team, made from volunteers, is creating value matched to similar professional teams. Since we are non-profit oriented and we don’t sell anything, we don’t make profits and we don’t match salary costs either. For one(1) general IT person (not kernel hacker) in my country, they start at around 35.000 EUR / year (taxes included). In Germany, this is about doubled and in Swiss tripled. Life expenses are in a similar relationship. Plus there is company running costs. Thanks for @Jon Smirl for data from US.

    We can be happy volunteers or just another precarious workers. Now do the math.

    @JotaMG
    We don’t want to employ slave labour, which is only possible within 1% of this money and I wanted to emphasise that. If we need to pay for people and/or teach them – students – this is yet another long term investment, which is not exactly helpful, when you need a full blown workforce to relieve stress on the current team.

  32. Jerry
    July 16th, 2017 at 21:13 | #32

    @Varghese
    Yes, RPi is still a better choice if you need open source 3d graphics and HDMI CEC on your NAS. It’s sometimes hard to choose between gigabit LAN & native SATA ports and a 100M LAN shared with USB/SATA bridge on the same USB2 host.

  33. TC
    July 17th, 2017 at 04:13 | #33

    Igor :
    For one(1) general IT person (not kernel hacker) in my country, they start at around 35.000 EUR / year (taxes included). In Germany, this is about doubled and in Swiss tripled.

    Munich, right? πŸ˜‰

    Here in the North of Germany IT’lers are more near to the 35k, tho…

  34. Eric
    July 17th, 2017 at 05:42 | #34

    @Igor

    I just hope JotaMG is not an EE or softdev teacher, because that’s just scary in terms of emptyness and basic financials…

    beat armbian for 240$… he should start a freakin kickstarter!

    If I would be half as bad as trolling, I’d say he’s from the “if you can’t do it, teach it”. Oops.

    Jealous people seeing 5 didgits going boohoo is sad. 24K is nothing in terms of operations, its not like if Armbian won the lottery nor its a “waste of money”, because one of the reason I (and a lot of people) tried a few orangePis is because of armbian support.

    So @JotaMG please go back under your bridge, light a fire in that barrel and ponder…

    Then, start coding some awesome stuff and make the world a better place through some hard work, not harsh words, and only then you will earn some credits and enough respect to piss on any open source projects and call something a waste of money, because right now, the only thing that looks like a waste of money is your salary and my time replying to your statements.

    Keep it up armbian!

  35. July 17th, 2017 at 10:06 | #35

    Orange Pi Plus 3 (top left board in photo) looks to be interesting. Upgraded from 32-bit H3 to A64, and if it has 3 or 4gb or RAM then yes, virtualization becomes possible. SATA speed will probably be limited, but, we’re not expecting to break any speed records at this price point. πŸ˜‰

  36. July 17th, 2017 at 10:06 | #36

    Updated post with feedback from Steven, which I added at the end of the post.
    Only some boards are sold at BoM costs, and the government only help with parts of the costs involved.

  37. nobe
    July 17th, 2017 at 10:31 | #37

    hopefully the useless internet drama will now stop…
    @cnxsoft
    very interesting feedback, thx.
    also this is close to @lvrp16 ‘s comment.

  38. tkaiser
    July 17th, 2017 at 12:22 | #38

    @miniNodes
    That’s no OPi Plus 3 but you’re talking about already selling ‘Orange Pi Win Plus’ (A64, 2GB DRAM, 4 USB ports sharing bandwidth). IMO way more interesting is the Plus 2E successor called ‘OPi Prime’ with H5 that is two times on the picture. Both boards have 2 GB DRAM since this is the maximum we get with Allwinner anyway (won’t change with H6 too unfortunately).

    ‘SATA speed’ does not exist since these things have no SATA. But when using a good UAS (USB Attached SCSI) capable USB-to-SATA bridge performance is already known and can be found as always in Armbian forum: https://forum.armbian.com/index.php?/topic/1925-some-storage-benchmarks-on-sbcs/&do=findComment&comment=34192

    The only interesting boards I spotted is the large one based on RTD1295 with native SATA, dual GbE, dual HDMI, 6 antenna connectors, USB-C, at least another USB3.0 port and 2 x mPCIe (but no one knows about software support yet and I’ve not seen any performance tests for SATA and GbE here). And then on another picture the small H3 or H5 OPi Zero board which seems to expose Fast and Gigabit Ethernet (EMAC/GMAC) at the same time.

  39. July 17th, 2017 at 14:31 | #40

    @cnxsoft
    Their first “Advanced CPU module” will be based on Rockchip 3399 processor. No price yet, except it may be close to $100.
    RTD1295 board is still being worked on, and not quite ready yet.

  40. Mum
    July 17th, 2017 at 15:13 | #41

    To all this debate about funding Armbian…

    Of course it’s nice that the project gets money so they can improve build system/infrastructure, etc. As Igor mentioned, it’s structured as non-profit, so money just increases the capacity. No one is contributing to Armbian and expecting to make money from it.

    To everyone else saying $24,000 does not pay the salary of employees, of course you’re right. But this has never been the point of open source. The Linux kernel is arguably the best example of this. Sure, a lot of people who commit to Linux work for companies like Intel, AMD, ARM, etc. but you as a normal person without any help can commit a patch/feature to the kernel. Open source is mostly about people volunteering in their spare time.

    I hope Orange Pi follow through and donate money to Armbian, since they are getting a lot of benefit from the project. In addition to this, it would be great if they listened to the community during design (e.g. don’t release products with crummy WiFi radios), and also worked more to distribute better OS images (unlike the BSP garbage they give you now, which only harms their reputation because everyone thinks the boards needs a fan and performs like trash because cores are too hot and shut down).

  41. tkaiser
    July 17th, 2017 at 15:32 | #42

    @cnxsoft
    Yeah, RTD1295 Ethernet performance looks promising but we have 2 GbE ports here and iperf/iperf3 are no real network benchmarks anyway (as quick ‘saturate link’ test sufficient, for everything else rather useless).

    Wrt storage you can’t test with HDDs anyway (since way too slow). If you want to test for SATA 3.0 bottlenecks you need something like a Samsung 850 Pro with at least 256 GB capacity (since the 128 GB variant is still too slow: write performance is below SATA 3.0 limitation). Wrt USB 3.0 the same applies (USB3/UAS is good for up to 400 MB/s). With a connected HDD you can tell ‘fast enough for home use’ but that’s not really a benchmark describing the device’s capabilities/limitations.

    Wrt random IO performance that usually depends on the storage device connected but there are also host limitations possible. To get high random IO performance you need features like native command queueing (NCQ) working on both sides of the connection (so here ‘USB Attached SCSI’ — UAS — becomes a must when talking about USB storage) and with both SATA and USB host controller limitations might need some tweaks/quirks.

    For example with some cheap mPCIe attached SATA controllers like ASM1062 (the one on Banana Pi R2 for example) you run in troubles with some kernel/settings combinations and sometimes it’s even necessary to more or less make NCQ useless by manipulating /sys/block/sdX/device/queue_type (none, simple, ordered) or play around with /sys/block/sdX/device/queue_depth. This can horribly impact random IO performance especially when there’s more than one disk used in parallel. Is that something we have to take care with RTD1295’s SATA implementation? We don’t know yet since no one looked into.

    None of this stuff gets addressed by SoC or board makers and that’s where Armbian comes into play. We invest an insane amount of time and efforts to test for this stuff, come up with better settings, identify and eliminate bottlenecks. These settings if proven to work correctly then become part of the automated build system and after some time appear hopefully everywhere so every SBC user can benefit from regardless of the chosen distro.

    At least that’s how it’s supposed to work! Wrt board makers there are some like Xunlong who listen, adopt community’s work and ideas, help the community by providing sources, correct documentation/information and — not that important — free board dev samples (I can buy boards easily myself, but I can’t buy sources and correct information). And there are of course some other board makers who behave like ignorant brain damaged retards.

    And even on platforms Armbian hopefully will never support directly users can benefit from the work done at Armbian after some time, eg. Raspberries: https://forum.openmediavault.org/index.php/Thread/18991

  42. JotaMG
    July 17th, 2017 at 17:20 | #43

    I say it again: giving money to Armbian is a complete lost of time and resources.

    ( why? well, you know… because money just can’t buy talent and inspiration… )

    And Igor comments here (and Thomas Kaiser on other topics) pinpoint with extreme precision why Armbian is going nowhere: you now feel like gods of the arm world !

    • Daniel
      July 17th, 2017 at 18:38 | #44

      You are aware that you and your students are perfectly free to create something better than Armbian, aren’t you? Who knows, if anyone would use it, they might even pay you.

      Every Armbian dev contributed their time and knowledge to create what is objectively the best OS for the supported SBCs, so they can rightfully voice their opinions and experiences, whether others agree or not. I’d rather read tkaisers complaints than buy an OPi Zero, or read about those NAS tweaks and apply them than not knowing about them.

  43. tkaiser
    July 17th, 2017 at 17:43 | #45

    Mum :
    it would be great if they listened to the community during design (e.g. don’t release products with crummy WiFi radios)

    Well, XR819 on OPi Zero is one example for a bad Xunlong hardware choice and GL830 USB-to-SATA bridge on their first H3 board ever (Orange Pi Plus and later ‘Plus 2’) is another. Do you know a third?

    2 years ago GL830 was used everywhere and is still used on some new products today but at least Xunlong doesn’t use it on their new designs any more (on their NAS Expansion board there sit 2 great JMS578 USB-to-SATA bridges and I really hope the small new OPi Zero variant with 2 Ethernet jacks that can be seen on one of the photos will be the perfect companion to this NAS Expansion board soon… and not just a prototype).

    Wrt XR819: Xunlong was the first who used this and most probably didn’t know what to expect. It’s ok-ish for the use cases an OPi Zero should be used for (IoT node) and we currently have a lot of users complaining that Armbian’s mainline images don’t contain support for it any more (since ‘works perfectly for what it’s made for’). The real problem with Wi-Fi on OPi Zero are users that only ‘buy as cheap as possible’ and then expect from a $7 device performing like a $100 device and getting free life-time professional support for free.

    BTW: FriendlyELEC also used XR819 on their first NanoPi NEO Plus2 samples but then chose to redesign the board this time relying on AP6212A for Wi-Fi (and also a better power design since they learned from conversation with community that with mainline kernel voltage regulation works unlike with Allwinner’s shitty H5 BSP where it’s still impossible to adjust CPU core voltage). In other words: Xunlong’s use of XR819 saved other vendors from making the same mistake πŸ˜‰

    There are rumours that most recent OPi Zero PCB rev 1.4 has a different power design allowing XR819 to ‘perform’ better. For reasons unknown to me nobody is testing this and I really don’t feel like wasting my own time again to do such tests in my spare time.

  44. Jon Smirl
    July 17th, 2017 at 19:59 | #46

    @tkaiser

    Maybe Armbian could help out on the Android front. Right now Android support is chaotic. A first step would be to simply get the releases we have for Allwinner and Rockchip checked into source control and then verify that they build the images without issues. Doing that would provide a central repository for people to work in since the vendors won’t supply one. It also eliminates the need for everyone to painfully copy 6GB files out of China.

    Doing this generates a significant git load so maybe do it on github or gitlab and then just manage it via Armbian. I am using gitlab since they allow larger files (zipped toolchains). Also note that you do not need to mirror the complete AOSP source, you only mirror the repos Allwinner/Rockchip changed. For A64 this is about 10% of full AOSP. I can help set this up, but ayufan knows more about how to automate it all.

    Long term purpose for doing this would be to get these Android releases running on mainline kernels that can be updated. Plus maybe Allwinner/Rockchip might even let us mirror their secret git servers after they see this.

  45. Drone
    July 17th, 2017 at 21:05 | #47

    @theguyuk
    @theguyuk, you said, “Which countries factories you thinking of?” Answer: Obviously any factories EXCEPT Chinese factories!

  46. TC
    July 17th, 2017 at 23:05 | #48

    tkaiser :
    There are rumours that most recent OPi Zero PCB rev 1.4 has a different power design allowing XR819 to β€˜perform’ better. For reasons unknown to me nobody is testing this and I really don’t feel like wasting my own time again to do such tests in my spare time.

    is there a 100% chance getting a 1.4 rev when ordering now?
    probably just not enough people got one so far…

    irt to what you do for the community, 24k is a joke, tho.
    Armbian is like the actual SW dev department of Xunlong

    @cnxsoft: can’t you just ban that JotaMG troll?

  47. Tj
    July 18th, 2017 at 01:38 | #49

    Still dreaming of cheap soc with multiple native sata ports with minimally 4gig ram.
    Dreaming to replace linux with zfs on linux with arm based.

    My mutiple zfs are running x86 entry level server board

    My dream is minority voice in cheap arm sbc board on my understanding.

  48. tkaiser
    July 18th, 2017 at 01:48 | #50

    tkaiser :
    And then on another picture the small H3 or H5 OPi Zero board which seems to expose Fast and Gigabit Ethernet (EMAC/GMAC) at the same time.

    In the meantime Icenowy fortunately reminded me that’s not possible since the only Allwinner SoCs having both EMAC/GMAC are A20 and R40/V40/T3. Looking at http://farm5.staticflickr.com/4280/35919268085_43dbf1ec0c_k.jpg again another variant would be H3/H5 combined here with the usual external RTL8211E PHY (7x7mm 48-pin QFN package) using the SoC’s internal Gigabit Ethernet MAC and one of the SoC’s USB2 host ports (usb1) directly connected to an RTL8153 USB GbE adapter (also 7x7mm 48-pin QFN package).

    Would be a smart choice since with mainline kernel we get +350 Mbits/sec with RTL8153 connected via USB2/Hi-Speed. So really looking forward to this new OPi Zero variant combined with Xunlong’s NAS Expansion board.

    @TC
    IMO if you think someone is behaving like a troll ‘don’t feed the troll’ should be the best approach. Everyone is free to express his opinions and banning/censoring/supressing is never the way to go. No idea which OPi Zero PCB rev you now get but based on user feedback when ordering through Aliexpress/Taobao it should be at least 1.4.

    @Jon Smirl
    We talked about that already 11 months ago and while a lot has changed in the meantime Armbian being focussed solely on Linux has not. Also Armbian ‘core team’ is still way too small for doing additional work. Let’s wait until ayufan is back from vacation and then maybe discussing this over at ROCK64 IRC?

  49. cortex-a72
    July 18th, 2017 at 02:04 | #51

    because SBCs have been made with mobile SoCs. that’s why your dream is so unrealistic. in addition to the “at least 4 GB of RAM” and “cheap” don’t mix well.

  50. JotaMG
    July 18th, 2017 at 02:40 | #52

    @Jon Smirl
    Excellent idea!

  51. July 18th, 2017 at 17:23 | #53

    good discuss , I learn more, anyone to do open source , it mean a dedicator, so , let everyone do the best they can do ,and help others , and respect the work of all people. The open source community do not need to argue with the attack. just do we can do .:)

  52. Jon Smirl
    July 18th, 2017 at 19:04 | #54

    @tkaiser

    For first step I would propose nothing more than getting existing Android releases checked in and building images. No need to write any code other than something like those automated build scripts Ayufan uses.

    The goal here is to have a single place where all versions of Android for Rockchip and Allwinner can be pulled from. You can make the git traffic almost disappear with an initial install script which first clones everything from Google’s AOSP server, then adds a remote to the Rockchip/Allwinner server and only fetches the deltas. With more work the repo tool could be modified to do this for you. Even without this change 80% of the traffic ends up on the AOSP servers. With the change you can probably push it to 95%.

    The big win will be in teaching the guys in China how to use this repo. Once they learn the procedure for copying a new release into it they can push only the deltas instead of 6GB files. They will only need to do that once for each delta and then their file hosting load will be gone. This procedure can be made very efficient in a place like Xunlong since they could set one git remote into the secret Allwinner/Rockchip server and another into the Armbian one and then do a pull/push procedure.

    Obviously getting direct access to the secret Allwinner/Rockchip git servers would be the most efficient since then we’d only need to fetch the deltas and not have to go through a pointless, error prone intermediate step.

  53. LinAdmin
    July 18th, 2017 at 22:21 | #55

    cnxsoft :
    Updated post with feedback from Steven, which I added at the end of the post.
    Only some boards are sold at BoM costs, and the government only help with parts of the costs involved.

    That was only one of the points I doubted…

  54. LinAdmin
  55. tkaiser
    July 19th, 2017 at 01:29 | #57

    @Jon Smirl
    Please check #ROCK64 channel backlog from today at irc.pine64.uk

    I quickly talked to ayufan, he meant that he’s happy with RK soon pushing to his own repos so wrt Rockchip IMO there’s not that much to do. With regard to Allwinner I would propose to simply forget about them. After their most recent R18 tinalinux code vomit I finally gave up on them. They’re obviously not interested in Open Source and even prefer board makers that do the same — see exclusive access to R40/V40/T3 for a vendor don’t giving a shit about providing correct information, documentation, schematics, sources and also not willing to fix security or instability issues: github.com/BPI-SINOVOIP/BPI-M2U-bsp/pulls still pending. That’s just INSANE!

    For my use cases all Allwinner devices would be only paperweights if there wouldn’t be some awesome communities doing Allwinner’s work (linux-sunxi, the *BSD folks and some others). But that might be just me being somewhat concerned about having to run smelly outdated kernels containing more exploitable security vulnerabilities than features.

    For me some Allwinner devices made by FriendlyELEC, Olimex and Xunlong (Orange Pis and the original Banana Pi) are still worth a look for low-end servers and especially inexpensive IoT devices (and the Pinebook for obvious reasons). With everything else coming from Allwinner I won’t waste my time any more (might change with upcoming H6 maybe if Allwinner chooses the board makers they partner with a bit more careful in the future and software support situation isn’t too horrible) and if I want to move away from Linux a bit more in the future the direction is *BSD and not Android. In other words: sorry, of no help here.

    We had the discussion here a few weeks ago already but back then focussing on board makers: if community constantly tries to compensate for their mistakes neither end users nor manufacturers will get what’s necessary (end users looking somewhere else, manufacturers improving). Why not dealing the same way with SoC makers?

  56. July 19th, 2017 at 02:17 | #58

    @tkaiser
    That’s also Android fault focussing on old and outdated kernels. And Allwinner will follow Android recommended stable kernels. They won’t care much about people wanted to use their chips for anything else.

  57. tkaiser
    July 19th, 2017 at 02:35 | #59

    @zoobab
    Sorry but https://android.googlesource.com/kernel/common/+/android-3.10 is not what Allwinner is using. Google’s Android 3.10 branch will receive a lot longer security updates than Linux’ 3.10 LTS kernel (but I still hope @willy doesn’t drop 3.10 maintainership in October πŸ™‚ ).

    It’s 2017. SinoVoip is about to release something called ‘Banana Pi M2 Magic’ soon. All available software offerings based on a 3.4.39 kernel (not kidding). Is this just INSANE or should this be called differently?

  58. July 19th, 2017 at 16:36 | #60

    @cnxsoft

    High Silicon Hi3798CV200 CPU based SBC (Single Board Computer) is also in preparation. πŸ™‚

    Android 7 ATV OS based + simultaneous boot Linux (AoL) + voice recognition function support.

  1. No trackbacks yet.