Interview With The Creators Of CHIP, A $9 Single-Board Computer

Single-board computing is hot on the DIY scene right now and riding that knife edge is C.H.I.P., a project currently in crowd-funding which prices the base unit at just $9. I was happy to run into the crew from Next/Thing Company who developed C.H.I.P. They were happy because, well, the project’s reception has been like a supernova. Right now they’re at about $1.5M of their original $50k goal. We spoke about running Linux on the board, what connectors and pinout headers are available, as well as the various peripheral hardware they have ready for the board.

$9 But…

Last week I published a post discussing the possibility of Linux evolving into mainstream because of engineers growing up with boards like C.H.I.P, Raspberry Pi, BeagleBone Black, etc. At that point I was thinking that most of these boards would be used as a desktop computer. My thinking has changed a bit (which I’ll get to in a moment) but I think it’s important to note that most people will not use the $9 base model for this purpose. It only comes with composite video. So the “but” about the price is you need to spend an additional $10-15 for VGA or HDMI adapters. As I said before, that’s not a deal-breaker.

C.H.I.P. $9 Computer ModuleHaving now seen it in person, I’m beginning to wonder if this won’t be used for a lot of headless projects?

…yeah, maybe just $9

There are a ton of reasons this should make it into projects that don’t have any need at all for a display — giving you the savings of not taking up board space or BOM cost for unused ports.

It has power-management (X-Powers AXP209 PMIC chip) designed into it, allowing for easy operation and charging of batteries. There is WiFi and Bluetooth, camera support, 8 GPIO, and your standard UART/SPI/I2C. Think of all the relatively heavy-lifting embedded projects that are begging for this horsepower and pricepoint: rovers, drones, and visualisation displays (think huge LED matrices or giant flip-dot displays)  to name just a few. I’ll be interested to explore the latency of the GPIO when I can get my hands on one of these; an issue we’ve heard about with Raspberry Pi powered devices.

DSC_0842You Can Take it With You

I’m not quite sure what to think about the Pocket C.H.I.P.

It’s a portable-form-factor which has a huge screen and a full keyboard. The base unit plugs into the back, inside of a case with an interesting geometric shape (anything is better than rectangular, right?). Of course it all runs from battery.

I don’t see myself using something like this but I’m obviously not the target demographic. In the age of smartphones it’s hard to envision something that’s not dirt-cheap taking hold. On the other hand, just having the reference design as an example of what you can build up around the base unit is a great marketing move on their part. “Hey, look at this ARM chip on a PCB” — meh. “Hey, watch this ARM-board power a portable touch-screen computer with clicky keyboard” — Awesome!

125 thoughts on “Interview With The Creators Of CHIP, A $9 Single-Board Computer

  1. For me the power of the Pocket C.H.I.P. is to do the initial setup of your headless C.H.I.P. I can do my install and then wifi setup and then unplug it and move on to the next one.

    1. I never plugged a screen or a keyboard on my BBB… Most of the time on embedded (or not) computers, you only need ssh or serial. And if you ever need to use a screen, you have X forwarding via ssh (which is really cool!).

      By the way, to make a “drone”, what’s the most efficient solution (in term of lag) to stream video from a webcam?
      Webcam are not block devices, so no NFS, cat can’t read it (so no netcat), X forwarding is much too slow, video players are too heavy…

      1. IMHO the most efficient solution is to send a compressed video stream over the network. There are several options for both ends.

        The difficult bit is compressing the video at the source. On Allwinner chips there’s a hardware encoder, however nobody (to my knowledge) has produced a solution that’s faster than compressing it in CPU. Other ARM chips, particularly if they’re tablet / phone / set-top-box oriented will more than likely have a hardware encoder too.

      2. Thanks. Sending compressed video over network is what I want to do, but webcams do not seems to output frames untill asked proprely by a program (UVS stuff…). As I said, cat does not catch anything from them (and give an error). For the “multiple JPGs”, that’s already what most webcams do, it’s called MJPG, for Motion JPG. To reduce lag, I would like to “dump” that on the network, just like I would do for anything else: with netcat, my problem is the software to “open” the webcam., as cat can’t read it and I would like to avoid re-encoding it as most utilities does. For the streaming part, I can use netcat or something that does RTSP (Realtime Streaming Protocol), but I haven’t tried the last one, as at the moment, I do not have something suitable to take video out of the webcam. Most “light” tools to do so can only do one frame at the time, like fswebcam.
        And the objective is to do that on my BBB, so no hardware encoder or fancy desktop processor.

  2. I love this thing. I hope they can meet all their goals and still make some $ for their fairly large team.

    The pricepoint is challenging seeming to me. Once its out of kickstarter I will be ordering one for sure.

    1. These things won’t be shipping until Jan-Mar 2016. I can practically guarantee you’ll be able to order a real product direct from China for about the same price (once you factor in shipping) before then.

      1. Their time frame opens them to a lot of rising competition, for sure. That is the main reason I didnt pledge.

        Honestly, though, I prefer to work with well documented development tools, so i generally only buy asian dev boards if I am going to use it to demo a specific chip, as opposed to using them as a module within a larger project or product.

        China cannot compete with real development tools unless they make it a real goal, like seeedstudio has.

      2. As far as I know this is from china. They have partnered with Shenzhen-based accelerator HAXLR8R. Which means that most of the production from individual components, parts, PCBs to P&P of the final product will likely be done in the same city.

  3. Mike! Mikey!! You didn’t ask the most important question: is it really 9$ computer is this just a miragé – the actual costs being hidden in the 20$ shipping fee? Can they actually really really promise be selling these things for 9$ in future – without embedded costs (pun partially intendeded).

    1. There isnt a “20 dollar shipping fee”.

      I placed some of the items in the kickstarter checkout and none had more than $10 shipping to a place in USA. Might be different international – but then, inst it usually?

      1. Exactly. Depending on where you are in the world, the Raspberry Pi (any version) can cost up to its selling price in shipping fees, easily doubling what you pay for the hardware itself. That doesn’t change the fact that it’s a $35 computer. Just goes to show that even with a $9 computer, some folks simply can’t be satisfied.

        Probably the same people who 10 years ago complained that the completely free Ubuntu discs being shipped to them anywhere in the world for free, took longer than six weeks to reach them.

          1. Chinese e-packet shipping is dirt cheap, and E14/Farnell’s shipping is probably more expensive – that and their ‘handling’ charges are bad. Also many Pi’s aren’t even made in China, but at a Sony plant in Wales.

            But as for the board itself, the RPI foundation had a lot of teething pains that have since been sorted out. If they had Kickstarter, it might not have been quite as bad – one of their mistakes was *seriously* underestimating the demand, the second was Chinese production teething pains. They eventually got past those by passing on production to Farnell and later Sony.

    2. I agree that $20 is a lot for shipping, but guess what. Shipping is tough and it costs money.

      Anyone who has ever sold a lot of something knows that losing customers to perceived high shipping costs is an issue. Because of that I’m sure CHIP will do whatever they can to drive the costs down.

          1. The US government DOES subsidize their mail system. Decreed by an act of Congress, no less. That’s why USPS is cheaper than other services such as FedEx and UPS for the same shipping “products”.

        1. China and US economic leaders are, apparently, working to achieve labor quality and cost parity.

          As a result outright economic warfare against USA markets is tolerated – for instance, china heavily subsidizes shipping costs for their ebayt\amazon\alibaba economic attack army.

          Normally a country will protect its domestic markets with tariffs, vats, etc. But not USA and China… USA lets china debase our markets openly.

          You would know all this though, I would think, seeing as you compete directly with arduino.

        2. Chinese manufacturers have an advantage in shipping into the US. They can ship a package into the US for less than half the cost of me shipping a package of the same weight into the next city.

          The 2010 ePacket deal has lost the US Postal service millions. And makes small-scale commerce between the US and China a one way street due to shipping cost.

          https://www.uspsoig.gov/sites/default/files/document-library-files/2014/ms-ar-14-002.pdf

          http://www.washingtonpost.com/news/storyline/wp/2014/09/12/the-postal-service-is-losing-millions-a-year-to-help-you-buy-cheap-stuff-from-china/

          http://www.ecommercebytes.com/C/letters/blog.pl?/pl/2012/4/1334191984.html

          http://community.ebay.com/t5/Shipping-Returns/Why-is-China-to-US-shipping-cost-so-much-cheaper-than-US-to/td-p/2838826

        3. They can always ship the lot to China and have to ship there in those fulfillment centers. Since they have all the paying customers all up front plus a shipping time of next year(?), there is no excuses not to do it that way.

          1. There are some valid reasons, if not excuses, for a organization to want to handle the storage, testing, packaging, shipping, etc. of their products.

      1. The HAD store is $20 international shipping. That $20 price tag sticks so I never go looking in the store even though I would like to be a supporter.

        The smart thing to do would be to mix it up a bit with the shipping costs. Then people might go and look at something to see what it may cost instead of not bothering because they remember that the shipping is $20.

        The old sales model of making a profit out of a product is dead in the water now and it definitely wont work for you because the expensive shipping costs will prevent the ‘first sale’. This is not a lost sale, it’s a lost customer.

        The model of making a profit out of a customer is a lot more successful because it allows you to make a little less on that first sale and pick up later.

        If you seriously want to take on international sale then you have to cut margins for international customers as they will not sympathise with you and pay extra for your inefficient postal costs.

        Have you thought about a member discount? They are more likely to be ongoing customers. What do the members want in the store? Perhaps ask them.

        People need to sample your service first because they won’t pay higher to premium prices (compared to international competitors) for a service unless they know the service is good. It’s not just about the product.

        Having said all of this I should mention that I rarely buy from the US. I bought a Papilio One and LogicStart Megawing as a bundle on special for about $120 USD and with free shipping. I bought some retro 4164 DRAM chips from Jamico because I literally couldn’t find them anywhere else on the planet.

        It’s very annoying that the US postal industry is killing the export of electronics ie small products. I quite often end up with fake chips from China as a result of avoiding your shipping costs but even so, I still come out ahead. I have become very good at noticing fakes.

        Anyway that’s my 2 cents worth. I find the whole US shipping thing frustrating (not just the HAD store) because it’s cheaper for be to buy small quantities if chips ‘Made in USA’ from some other country. I mean – just how ridiculous is that really?

          1. Because the comments system is still broken (for edits and comment nesting – and I will say once again that there are many plugins for this for WordPress that support anonymous style commenting, almost no “budget” required) and Mike cannot be replied to, your store says:

            “Shipping costs:

            USA: free on orders $25+, otherwise $5.

            Canada: free on orders $50+, otherwise $12.

            International: free on orders $50+, otherwise $17.”

            To buy a small esp8266 for $6.97 it would cost me $17 in shipping (to Nuuu Z-land). On Ebay the exact same small size esp8266 module costs $3.06 with free shipping. That’s almost an 8:1 cost ratio.

            Economics !== disrespect.

          2. Dear Fennec,

            All we can do is “negotiate with reality” and find a local store which provides the same (or equivalent) goods at competitive prices. For example, the ESP module I got, cost me ~$6, and because me and some other folks bought some more stuff we haven’t paid for the shipping at all. We haven’t supported HaD but we supported a local vendor, who brings lots of nice stuff and 2-3 day shipping not possible at any reasonable price across the Pond.

          3. Reply to Fennec:
            NO KIDDING?! It costs less to ship from China to NZ than halfway across the planet? OMG HaD is gauging us!

            HaD is not set up to massive amounts of international shipping like most Chinese Ebay stores are and does not get the price advantage in slave labor that China has. That you are ordering one inexpensive component does not change the fact that it has to be packaged and shipped. Packing materials will cost the same, labor costs the same. HaD had to buy and import the parts FROM China so no crap the Chinese have a pricing advantage. And shipping costs more out of the US than China. By your own admission, international shipping is free for orders of $50+. That seems more than reasonable to me since most companies here only offer in-country free shipping if you spend way over $50.

        1. It is not even going to be $9 for the final product. There are a lot of overheads: Kickstarter/bank cut of handling the money. Then you factor in the R&D cost – parts/prototype revs/(compliance testing?), setup charges, shipping/handling cost of the components etc.

          Even if these guys are not paying themselves, they are also going to have fun with the tax man and having to deal with lots of paperworks.

    3. There is very easy way to solve this – just make a 10x CHIP module only reward for kickstarter. There is one for PocketCHIP, how come there is no for CHIP only? because to me it looks like it was easier to hide additional costs in PocketCHIP price than to deliver 10x CHIP boards with some ridiculous (even more than 20$ shipping fees). Also guess what? 10x PocketCHIP shipping costs 100$ WTF? Are they paying for the duty&taxes?

      This stinks. I would love to see cheap linux SBC for all to use, but with scamy looking shipping fees and ungooglable name I don’t see CHIP being it.

      1. 10x CHIPs?

        Screw that, I’d be happy with *two*.

        It’s pretty clear that they won’t do that, though, meaning they probably can’t make a profit at a $9 price point. Virtually all the Kickstarter comments are about adding more CHIPs, or a multi-CHIP tier, and there hasn’t been one comment from the creators about this.

          1. It’s not really any of those. It’s just a loss-leader, although the “loss” here might be just in profit, rather than actual “loss” due to selling an item below cost.

            Everything *above* the CHIP level is overpriced. The HDMI/VGA adapters are way expensive for ~10K volumes. How do I know? Because the VGA adapter is just the cheapo Olimex circuit: https://olimex.wordpress.com/2012/06/12/low-cost-lcd-to-vga-adapter/ . Which you can see in the Kickstarter pictures. The total parts cost of that board is probably $1.50.

            In 10K volumes, this is not a $10 board. You could *make your own* for less than $10, which is what I’m planning on doing. Their cost for the VGA board is probably $2-3, at most.

    4. As far as I know they are only obligated to deliver the product promised for a certain pledge amount or return the pledge. Not under any obligation to continue selling the products for that amount. You mist be a wealthy individual if you can make that much by imbedding manufacturing costs in a shipping charge of $20 for a $9 product and still deliver that product.

      1. As well as the extreme culture/behavioural changes required to be a first world country. People neglect this and think that there’s an easy way out, get all offended when “destroying” a “race’s” culture is mentioned. Whereas really, forgoing all sorts of unnecessary stuff like that is, well, necessary. Just look at the steady decline of religion, one of our last cultural aspects, in Western countries.

          1. Yes, but education results in improved infrastructure but also a decrease in culture/traditions. There’s less desire for inefficient and unnecessary conscious traditions and so they are replaced by unconscious habits.

            The dynamics occur when immigrants move from their own cultural/traditional country to a Western country, for example. Perhaps one of the reasons we forgo these things is because intelligence drives us towards being a unified humanity, without creed, and culture equals creed which is divisive.

        1. Religion is a personal freedom, but at the same time IMHO decline of religion is a good thing. There are far too many monstrosities committed by man in the name of religion even today.

          I also do not like the recent trend of messing up the education system to accommodate religion.

    1. Those cheap Chinese “smart phones” or tablets are the 3rd world computers. They combine a connection, storage, input device, display, case and power source in a single device.

      So far RPi and other SBC would need additional components just to make a usable thing. It is unlikely they run the “computer” as a headless embedded controller. So even the C.H.I.P. would cost multiple times the $9 amount.

  4. I wonder how much they have to pay in patent royalties for the right to use OFDM in the wifi (just because the chip maker pays a royalty doesn’t mean you don’t have to too)

  5. The RPi B+ was recently reduced in price to $25, and shipping to most western countries is $5-$8, whether you order 1 or 10.
    For headless applications where Mike says CHIP is a better fit, I’ve previously pointed out that RT5350F-based router sticks are available now for <$8 including shipping. The one I have has 2mm pads for UART which can also be used as a couple GPIO. A couple more GPIO are accessible on pads near the USB connector. If you need low-latency GPIO, plug a digispark clone or a USBASP into the USB port – you can get them for $2 shipped on Aliexpress.

  6. I love the pocket chip to death but all i can think is id like to plug a Pi into it :P

    My only issue is with an early next year shipping date… i cant help but feel a better Pi or Arduino will be out by then and make it not that worth it.

  7. Ah, hoped to get some questions about C.H.I.P. answered by this interview but no luck. Like for example, are the GPIOs 5V tolerant? Will they add an unregistered mail shipping option, or alternatively ordering many boards with a single $20 shipping fee? The shipping fees right now completely cancel out the price advantage of this board.

    I’m surprised no one is asking the interesting questions, neither in the video interview or on the kickstarter FAQ. I don’t believe the video output on CHIP or the BBB or the Raspberry Pi is used by even half of these boards’ actual users. On the other hand having cheap IMU shields, or fast and 5V-tolerant GPIOs would make these boards really useful for the types of projects they’re best for (home automation, drones, Arduino-killers, …) I can’t wait for someone to finally put my crap 8-bit $2 Arduino clones out of use, but so far none of these more advanced, “revolutionary” boards match them in terms of I/O due to something that’d would add less than $0.50 to their BOMs.

        1. Touche.

          Though, as someone who works with things like photomultiplier tubes for cutting edge uses, certain things, like voltages, remain a practical concern. (Sitting here looking at bias amplifiers and such)

        2. Do ANY of the regular parts you have tolerant of those voltages without voltage divider or level translation etc?
          Even the linear chips (for a while have resisted 3.3V for performance reasons) has families for 3.3V and lower rails.

          It had been more than about 20 years for 3.3V logic since around time of the old SDRAM/Pentium day, time to catch up with the rest of the industry.

          1. heh – the large values are for use on photomultiplier tubes, for particle detection – so it isnt into logic.

            I agree, voltage trends are changing – 5v may eventually be only for legacy parts and will be more or less ‘obsolete’ then.

      1. the thing is not everyone is looking at a dev board for digital only signaling. 8bit atmel chips (and yes ‘duino) have the 5 volt analog inputs and pwm outputs I need to work and interface with analog synths. I would love a $9 32 bit board with 5 volt in and out so I don’t have to spend extra on parts for level shifting on every single output and input. that drives the cost up beyond the cost of the original dev board. I’ve gotten to the planning point of using the atmel flashed to ‘duino as an interface chip for the analog side and am looking for something FASTER to handle some dsp tasks and housekeeping/storage. the chip might be it

  8. Excellent showcase Mike. Seriously these things sound pretty damn neat.

    Anyway, a heads up about Allwinner A13 some folks might be concerned about the bootloader, Mali400, libvp62, ffmpeg stuff.

    I “think” Allwinner addressed and fixed the way that they ripped open-source project code. (New repo commit 2 days ago) https://github.com/allwinner-zh/media-codec

    There was (is?) an angry wasp nest of folks: sunxi Allwinner GPL violations CedarX (A A20 project was cancelled because of it.)

    1. Allwinner has a pretty poor track record on GPL stuff and the linux-sunxi community can get pretty grumpy about it.

      That said, other than Mali and some other minor / unused parts of the SoC, I believe that the A13 chip (which the R8 is based on) can boot and run completely with upstream open source code, (Linux / U-boot) so the R8 should quickly be supported too.

      The biggest issue outside of Mali is that the developers of C.H.I.P. seem to have not contacted the linux-sunxi community at all which is pretty-much a prerequisite of having upstream support.

      1. Definitely agree with the comment about AllWinner’s track record. In my opinion, that will be this product’s demise into a short life.
        As for the cost, something is fishy there because if you total up a BOM cost even at some decent volume, you can’t make the $9 price tag even with a 0% profit.

      2. You all have done a good job of reading between the lines. My first question about the kernel availability was a polite way to see if they are talking about the issue. At this particular point I didn’t want to browbeat the issue.

        1. I’m on the linux-sunxi mailing list and this isn’t just “reading between the lines”, this is my inbox for the past couple of months or so summarised. =)

    2. I have not seen any news about allwinner actually fixing their failures of complying with GPL, only that they’ve tried to hide the fact that they are not complying. For that reason i will not touch allwinner stuff.

      1. They’ve just released GPL compliant code for the video decoder / encoder (though there’s an optional and apparently unused binary plugin) however there are still binary blobs for various parts of some of the newer SoCs. (A13 is fairly old)

        Details: http://linux-sunxi.org/GPL_Violations

        As far as I’m aware, for A13, (and probably R8) the only part which is both used and still closed source is Mali.

  9. At a price of 9$ for the board, how much does the processor cost?
    If it’s below 2$, why not build 50 of them on one board, parallel computing for less than 100$.
    Interesting times ahead, when 64bit + OpenCL Mali get in this price tag…

  10. Is this the new business model? Make your money on the crowd funding and open source a product with basically zero margin for future builders and destroy all the other board makers. Outside China anyway, which will change when they have their very very big economic adjustment one of these days.

    I guess it means doing a Kickstarter ever 6 months to a year for everyone. Hmmm. Kind of like doing each others laundry. I’m not sure I see a bright future in this development.

    1. This is one of those… “we’ll wait an see.” Give away the razors and sell the blades. Maybe even limit how many razors can be sold per customer. The real hustle would be hard coding the ASIC to only accept or communicate with whatever BS handshake encryption key they sneak in there. C.H.I.P. team might not even be aware of it happening.

      I am really interested in HPC/HMP computing, I am inclined to say the price point is REALLY attractive but given Allwinner’s track record *shrug* As it stands now seeing as how the semi-conductor industry is moving toward BGA it might encourage them to push toward maintaining QFP for their mid-range products.

      The ODROID-XU3 is a damn sweet product but *pssssh* the price point is WAY too high. (Projekt D3nv3r isn’t worth mentioning) Looking at other products say from Qualcomm, Broadcomm, Analog Devices (Blackfin 700’s) or even the wild menagerie of TI. You have to fully vest in their training and dev toolkits. (I learned while chasing I.T. certs this training and tools is a heavy revenue stream for these companies.)

      It almost seems like the best source of obtaining these high end chips with the least amount of headache (and maybe even cheapest option) to buy logic repair boards for cellphones and tablets to get the parts.

      Anyway the “big deal” now a days is who can get a affordable ARMv8-R (not A but R) out the door. Just for sh1ts and giggles ARM architecture is a giant mess.
      http://en.wikipedia.org/wiki/ARM_architecture#Cores

      Now researching the high end DSP/Video processing/GPU is entirely different hell. As [Julian Calaby] knows Luc Verhaegen has been taking a lot of crap both on the Lima project ( and Allwinner shills posting in the sunxi list “you should be nice Luc”.) Lima Tamil (Mali-TXXX) is a long road ahead.

      The 8xx snapdragon SoCs Adreno (freeduino) isn’t moving at a happy pace, too many pretedo-friends threadjacking etc. It’s like reading the “Big Brain Thread relating to the Propeller chip”.

      PowerVR? Hah! Imagination tech has zero interest in community. And the fact that some of the sauce was leaked in 2011 makes any Rev-Eng-ing project close to impossible, Microcode calls to the kernel to get thing working.

      Mediatek? just as bad or worse as Allwinner.

      Renessa has a R-CAR 2 chipset but it seems unless you are a automotive manufacture there is little to no details. Also not inventor/hobbyist friendly.

      Options as they are: thank you to Broadcomm for the (FINALLY) opening up VideoCore, anyone using vivantecorp IP (all their code is open from what I remember), and yeah we are stuck with Mali-400.

      I have spreadsheets upon spreadsheets that if printed out and affixed to a corkboard wall would send anyone running.

      My great hope is a future Teensy 4.x with a Cortex-M7F (hopefully with QFP).

      Finally, Stay away from trying to integrate Allwinner A80 into a project (“I don’t know what it’s called I only know the sound it makes when it LIES”). I’m sure there are folks in the Linaro group who try to be nice to all reps and companies involved in the Linaro project but that aside. The specs on the A80 versus actual performance is worse then nuclear waste.

      Anyway, This A13 project MIGHT be a good thing, force the hand of fab companies to make better documentation, and lower the price on components and improve the search and design tools.

      But in ALL honesty the life of a product is based on how much breathe, health and open the community is. Now the price points of said devices are a different story, Do I buy 12 Raspberry PI 2 or BeagleBone Blks? Or just a pick up some knock off Arudino Nano’s and 4 low end chromebooks. Hrrrmm…

      1. On the subject of sneaking encryption / lockout into these boards: Firstly DRM is expensive.

        The R8 appears to be a repackaged and updated A13 (looking at the specs on Allwinner’s site) so it therefore _only_ provides a parallel LCD display output. (Think of it as a digital VGA port) Therefore I suspect that the VGA adapter is just three parallel DACs and some analog components and the HDMI adapter is the same with a VGA => HDMI converter chip, so no possibility of hardware lockout there. (I’m not sure where they’re pulling the composite video from, incidentally.)

        As for the SoC itself, Allwinner’s SoCs boot like this: Hardwired code inits an SD (MMC) card and/or NAND flash and chooses one based upon some conditions (including the state of an external pin) to load a pre-bootloader from. This pre-bootloader (“SPL” (FOSS) or “boot0” (proprietary)) brings up the DRAM and some other stuff then loads the “real” bootloader (U-boot) which then loads Linux. This entire procedure is fully documented and with the exception of constantly changing DRAM controllers and multi-CPU considerations, has barely changed since the original A10. As far as I’m aware, no Allwinner based device has been produced which cannot been booted from 100% FOSS code.

        In terms of HPC / HMP, Allwinner SoCs tend to be clocked relatively low (around the 1GHz mark) so they wouldn’t be my first choice.

        My understanding of GPUs is that they’re hilariously complicated to get working and even more complicated to get working optimally – this is why the reverse engineered drivers have either come from people who are already in the GPU driver industry or are taking forever to get going. PowerVR adds a whole extra level of complexity given that (I understand) it can have different microcode based upon the application.

        At least Qualcomm, Broadcom (and nVidia) are actually providing some code to the community for their respective GPU solutions, and vivante appears to be sufficiently simple that a driver (“etna_viv”) has emerged from the reverse engineering project.

        My understanding of the A80 is that it was rushed to market (along with their quad-core A31(s) parts) and is somewhat unfinished. Upstream is still working on full support for it.

        Companies won’t produce better documentation unless it makes their lives easier. Linaro isn’t weilding a large enough bat to force their hand and they simply don’t care about the open source community. In the case of Allwinner, we have incomplete documentation for just about everything (datasheets mainly) but are generally missing programming documentation for some of the more complicated bits (E.g. Cedar (VPU), G2D (2D GPU) and DRAM controllers) – most of which have been reverse engineered.

        There’s a manufacturer sponsored community around almost every Allwinner dev board. (Like the Raspberry Pi forums.) Your average person doesn’t care about upstream as it’s not nicely pre-packaged for their consumption.

        1. No, the VGA adapter is stupider than that. It’s Olimex’s cheapo circuit:

          https://olimex.files.wordpress.com/2012/06/lcd2vga-sch.gif

          Look at the board on Kickstarter, you can clearly see that’s what it is. If my eyes are right, it has exactly the right number of components to match it. The Olimex circuit was for an A13, so their work is honestly completely done for them, right there. It’s trivial.

          Ditto for HDMI. The video out on the A13 is LCD, so converting to HDMI is just probably a TDA19988, which is set up to handle a 2-layer board layout if their connector layout is smart. So again, that’s a joke of a board.

          1. That’s exactly what I was saying, only slightly different.

            Olimex’s circuit is essentially 3 DACs, however I assumed they were a complete package not a buffer and resistor ladder.

            Based on that assumption, I assumed the big chip on the HDMI board was just a VGA => HDMI converter. (If you ignore the big chip on the HDMI board, it looks identical to the VGA board) If it is a TDA19988 or equivalent, it looks like it’s just a set of buffers which then feed the LCD lines directly into the HDMI chip. I also assumed VGA => HDMI chips would be cheaper than LCD => HDMI chips. I was wrong.

          2. The “slightly different” is a huge factor in BOM cost. The buffers are pennies, whereas a triple video DAC is $4-5.

            Since you’re starting off with a digital interface, going digital->analog->digital would be really impressively bad – a VGA to HDMI converter is *considerably* more complicated than a simple HDMI transmitter.

        2. Ew! They should put in a triple DAC if they want to sell 24-bits. You cannot get 8-bit DAC out of regular resistors as 4-5 bits is already stretching it. They should at least try to buffer the VGA video output with a $1 buffer chip. e.g. TSH344IDT. Some of the pin compatible lower bandwidth chips have built-in 3 poles filters.

          As for digital output, a TFP410 would be able to convert the LVTTL signals into DVI (which can be a subset of HDMI.)

          1. Well, the A13’s don’t have 24-bit video output – they only have 18-bit (note in the datasheet they have bits 7:2, 15:10, and 23:18). So the video already kinda sucks.

            But that board really does expose what their business plan is: the CHIP itself is a fairly complicated board – at $9 you’re never going to get a significant profit out of that. But the VGA and HDMI adapters are so simple, functionally you could take their cost to near-zero, and they’re charging *more* for those boards than for the CHIP itself. So you get tons of these boards out there, hopefully make enough from the VGA/HDMI/PocketCHIP bits that you can make it through Kickstarter, and then sell the C.H.I.P.s afterwards for $15.

            I mean, there’s no way they could end up trying to *sell* that VGA board for $10 afterwards and not expect someone to undercut them by half.

          2. @Rob

            I’m guessing by ‘regular resistors’ he meant probably 1 or 5% resistors. An 8-bit R/2R DAC ideally would require 0.1% resistors (since its LSB is 0.4% of the MSB) but practically requires even better than that without part selection (tuning the exact resistors) due to the aggregation of all of the errors at the major code transitions.

          3. @Pat,

            1% resistors will give a reasonably linear scale. Perhaps the whole scale is offset a little but the error does not accumulate in a way that strongly effects linearity.

            For example you mentioned that the error is accumulative. Well as you mentioned the total effect of a perfect resistor in the LSB is approx 0.4% of the total and so that resistor can never have more effect than this, so it can’t ‘add’ it’s 1% tolerance to the total. In fact its error as a percentage of total is 0.004%.

            So the error reduces with each R/2R from MSB to LSB and this in effect makes the scale more linear. The greater effect on total error (or offset) is the first two resistors.

            So if you are outputing something that is not sensitive to offest like VGA color or Audio the I would expect R/2R would be fine for 8 or even more bits. If however you are using it for quantified sampling then it’s not the best solution.

          4. FYI the circuit posted is not a R-2R ladder. It is a DAC with weighted sum, so you’ll be luck if you can find all the 2^n multiplier values on a log10 scale that the usual 1% resistors values are based on..

          5. @Rob

            The point is that, without extremely precise resistors, an 8 bit R/2R DAC does not give you 8 bits of resolution over the full dynamic range. Whether or not it’s “good enough” for a given application is a totally different question. And while crappy video can look “okay” in certain applications, it is really, really easy to make a video demo which demonstrates the limitations of the number of bits in video output. Any simple color gradient will *easily* show non-linearities in the DAC transfer function.

            “Well as you mentioned the total effect of a perfect resistor in the LSB is approx 0.4% of the total and so that resistor can never have more effect than this, so it can’t ‘add’ it’s 1% tolerance to the total.”

            This is wrong. You don’t have a resistor per DAC value, you have resistors per DAC *bit*. So some resistors are ‘more important’ than the other ones. Even for a weighted sum DAC (like the Olimex circuit is), the resistor tolerances obviously add if more than 1 bit is on. But for an R/2R DAC, it’s even worse.

            For an R/2R DAC, if *only* the high bit is set, for instance, you’re trying to have 15 resistors add up to *exactly* the same value of the high bit’s 2R resistor. If the tolerance of each one of those 15 resistors is 1%, then propagation of errors says that the combination of *all* of them has a precision of ~4%.

            And a 4% error in an 8-bit DAC corresponds to 10 DAC counts: or in other words, not nearly as good in terms of maximum error as a ‘perfect’ 6-bit DAC. Will that 8-bit DAC look better than a 6-bit DAC? Sometimes. But that depends on the image, and on the *actual* DAC itself.

  11. I was running a Linux Stamp board that has no video output, for about 4 years as a backup SSH server on my home LAN. It cost around £100 and replaced it with a Pi recently because it is far easier to keep it updated (better Operating system support).

    $9 does seem like a low price point but due to the fact that VGA/HDMI add on bumps up the cost and makes the physical profile bigger, a near equivalent might be the Raspberry Pi B+ for £16:-

    http://www.wired.co.uk/news/archive/2015-05/15/raspberry-pi-price-drop

    Having said that, I look forward to manufactures producing almost throw away Linux boards for a few bucks which will be used instead of the micro controllers that are common today – obviously massive technology overkill but that is the market that I see these types of computers trying to sell into.

  12. Composite video? Something (at last) to connect to all those portable DVD players in which my daughter has broken the drive mechanisms. She thought putting sand in one of them would make run better…

  13. Ahem…. back on topic. All of these little toys are nice, but they are just that, toys, until windows can easily be put on one and run unmolested windows apps without any issues. Linux is great, if you don’t want any software because nobody codes for Linux. It’s exactly why all these emulators were ported to the Pi but it took quite a while to get a front-end to actually launch them in a pleasing manner. That hobby level programmer that codes all the nice middle-ware just doesn’t exist on the Linux platform…. you have experts or people that don’t know what they are doing and really no one inbetween.

    Windows 10 is going to change all of this hopefully as it’ll apparently run on everything from Hal to a speak-n-spell.

    1. Um Linux accounts for about 35% of web servers on the internet. *nix like OS’s account for about 70%

      When you dig deeper towards DNS/FTP/SMTP/POP/Cloud servers you will find an even higher rate of *nix systems.

      As for coding for Linux! well just about any platform that runs on windows also runs on *nix, in fact most of them start on *nix and are then ported to windows.

      Apart from that you rarely need to ‘code for Linux’ … it’s all out there done already, just plug it in.

      By your definition, I am one of the in between people. I know enough to set up and manage a Linux server and manage security without coding anything specificity for Linux. My server coding is mostly CGI/PHP/SQL.

      There is nothing hard about Linux, it’s just different to windows and much more secure!

    2. What about the fact that Linux runs on the vast majority of the fastest computers in the world – and why? Because it is secure, optimised and scalable. There are no Microsoft offerings that come anywhere close!

      Windows 10 is a toy operating system that has none of the advantages of the greatness of Linux!

    3. > nobody codes for Linux

      Everybody codes for unix, except for windows coders — who only code for windows. To suggest that there’s more windows software out there than linux software indicates a lack of experience with real developers and real software.

      1. Nope. I once owned a fully kitted-out TI 99/4a that was used to run my local Radio Shack. It came to me with the expansion box that had a further 32k of RAM, a serial/parallel port card, and *TWO* 5 1/4 drives, along with serious business software like the Microsoft Multiplan spreadsheet that came on disk *and* cartridge. The whole thing cost more than $5000. The thought of a computer with more than a thousand times the power that costs more than two hundred times less? What’s to quibble over?

  14. Continuing the fake marketing that this is $9

    First impressions of this product: hipsters selling to hipsters

    Technically this cannot compete with raspberry pi as you have to spend a lot more just to get the function of a Pi B (which is $25?)
    This board + the HDMI add-on = $24 (+ $20+ for shipping PER BOARD)
    Also this has no microsd & a single USB

    The best thing about this project is marketing

    Smoke & Mirrors applies here,
    screenshots of a custom linux OS – no reference to sourcecode
    ‘fully opensource’ – closed source drivers (mali-400)
    ‘worlds first $9 computer’ – costs $29+
    (per board as they have no option to order 10 and pay combined shipping)

    Pocket CHIP – who has pockets that big ?

    Would suspect as these are made in china, shipping will be from china & cost nowhere near $20+

    1. You honestly described most of the shit they try to push here on HaD. I think we lucked out in that CHIP is not sold by supplyframe, or else there would have been 8000 ZOMG! posts like the arduino and raspi (fing hate that thing). It is all garbage to rip off middle class kids.

      1. Any suggestions for I/O ports extended from tablet?
        Is there a nice board for I/O ports from a tablet’s usb?
        Probably lots of people would go for an easy/economic solution.

    1. It’s USB2: Allwinner’s R8 SoC is a re-packaged A13, which is pre-USB3.

      Datasheet: https://github.com/OLIMEX/OLINUXINO/blob/master/HARDWARE/R8M/R8M_DataSheet_V02.pdf

      Look at page 9: in the pinout diagram, pins A12-A15 are the data+ and data- pins for both USB ports (host and OTG) there are no other USB pins, which means there’s no pins for the extra high speed lines of USB3.

      In general, people tend to advertise USB3 heavily, so if they say “USB” but not “USB 3”, then it’s safe to assume it’s USB2.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.