Odroid C2 Bests Raspberry Pi 3 in Several Ways

It’s been a big week in the world of inexpensive single board computers, and everyone’s talking about the new Raspberry Pi 3. It blows away the competition they say, nobody can touch it for the price.

Almost nobody, that is.

With a lot less fanfare on these shores, another cheap and speedy 64-bit quad-core ARM-based SBC slips onto the market this week, Hardkernel’s Odroid C2. And looking at the specification it seems as though the Pi 3 may be given a run for its money. Like the BCM2837 in the Pi 3 its Amlogic S905 SoC is a quad-core ARM Cortex-A53, but the C2’s 2GHz clock speed gives the raspberry to the 1.2GHz of the Pi 3. There is twice the RAM of the Pi 3 at 2Gbytes, and the onboard Mali-450 GPU can deliver 4K video.

Unlike the Pi 3 there is no wireless or Bluetooth on board, but the C2 has a Gigabit Ethernet port which is wired directly into the SoC. Compared to the Pi 3’s 100 megabit port which suffers through being on a USB interface, that’s likely to be very quick.

Storage can be a choice of either the usual micro SD card or eMMC. Given that the two boards share a very similar form factor it is no surprise that they have very similar GPIO capabilities, however it is worth noting that the C2 has a built-in analog-to-digital converter. As to operating systems, the C2 can run Ubuntu 16.04, or Android Lollipop.

Of course, we’ve seen so many boards touted as Pi-killers, and like all those also-ran tablets touted as iPad killers a few years ago we’ve never heard of most of them again after a brief moment of chatter. They look so good on paper but the price always lets them down.

The C2 could just escape that fate though, its $40 price point is very close to that of the Pi 3. Setting aside for a moment how much shipping and customs might cost for a package from Korea, that sounds interesting to us.

Why might you buy a C2 then, and why might you buy a Pi 3? That the C2 has a much faster processor is beyond doubt. This and its faster wired networking would make it a much more interesting prospect for anyone whose work involves network-attached data processing. But even though a USB wireless network adaptor can be had for only a few dollars the Pi 3’s onboard wi-fi and Bluetooth makes it much more attractive to a home user or someone using a computer on a platform unfettered by wires.

However impressive the C2 may be it is overwhelmingly likely that the Pi 3 will outsell it many times over. This will not just be due to the massive publicity advantage achieved by the Pi Foundation, but the huge ecosystem of hardware and software developers that have made the Pi boards perform to the limit of their abilities in all directions. If you don’t mind forgoing that support though, you could just find that the board from Korea gives you enough extra bang for your buck to make having it on your bench worthwhile.

We’ve followed the Odroid products from the start here at Hackaday. The C2 is just the latest of a procession of boards from Hardkernel, and we’ve featured a few projects that include them. Theirs is always the name at the top of the list when the subject turns to Raspberry Pi competitors, perhaps with the C2 they’ve got a winner.

Our thanks to [Derrick].

130 thoughts on “Odroid C2 Bests Raspberry Pi 3 in Several Ways

  1. I’ve tried Odroid products and they tend to perform. Also, the build quality is somewhat better than most of the other fruit-pi dev boards. But I’m sticking with the Raspberry Pi because any time I have an issue–someone out there has already solved it. That being said, if you’re in the states, the place to buy Odroids is http://ameridroid.com/

    1. If anyone picks one up from Ameridroid I would get the preloaded SD card or eMMC chip for the XU4/C2. I had kernel issues trying to put ubuntu 15.01 on a samsung UHS but debian jessie with MATE as the GUI works just fine.

  2. Forget the C2, though – look at the spec sheet of the XU4. Yeah, it isn’t in the same price point, but it doesn’t cost all THAT much more (well, at least if you’re some dude wanting to buy one and not an organization wanting to buy 100, I suppose), and it has an exynos double-quadcore processor and USB3 support. Still lacks SATA though, grumble grumble, but so it seems does every other computer-on-a-credit-card.

    1. The original Banana Pi and Banana Pi pro both supported Sata. Which is why I ended up getting a Pro, which also has built in WiFi but that really wasn’t a selling point for me. It was more about the potential for inexpensive bulk storage.

    2. The Olinuxino Lime supports SATA. As an added bonus it also supports a backup LiPo battery and a real power jack — the latter being something sorely missing from the Pi.

      1. I have (have had for a few years) a lime 1 (based on the A10) and the lime 2 (A20)

        and are the same cost as the raspberry Pi,
        and open sourced.
        and (like for like production dates) are faster than the pi, (e.g the olimex lime was dual core and faster clock speed than the pi.)

        I’ve never seem then even mentioned on sites such as hackaday before. – and that’s a shame that a more capable board with more pin outs, more built in peripherals, that runs faster never gets attention.

        (and that SATA is directly connected to the SoC not USB)

    3. Whenever someone mentions SATA on an embedded computer, I gotta throw in: you should consider hacking some variant of Dockstar or *Plug computer. I have a PogoPlug with 800mhz ARM chip and SATA, ethernet, and a few USB ports. It runs Arch. There are faster ones available too.

      1. Whenever someone mentions SATA on an embedded computer, the first thing I always ask them to check is how it is connected — many of those SATA ports are hanging off the USB bus.

    1. Software patents held primary by “MPEG LA” is the reason for some of the GPL violations. Which can probably be resolved in 2020, and beyond, when a large number of the current patents expire.

  3. That seems to solve almost all the limitations of the Pi3 that were mentioned in the comments on the other thread. Quite tempted to return my Pi3 and wait for this to become available. I’m sending rather large files across the network to my Pi3 and this would certainly speed things up quite a bit.

  4. The value of the Pi is the community, documentation, examples, etc. Technically superior doesn’t mean anything to someone if they don’t *need* the extra performance. But it’s always nice to have the option for those who can make use of it.

        1. Yes. I know. Documentation on that is also shy, but at least findable. But I got that to “work”. I just want to compare it to the other driver. As the VC4 driver is crashing my application, and shows a few visual bugs.

          I found documentation on the Pi to be far in between, and most of it is just basic Linux foo.

    1. The primary purpose of using a kernel like Linux is to abstract the hardware. So as long as the drivers are done the right way, the users should not feel bound to a specific board. As many point out, the GPU driver (or lake of it) actually broke that nice expectation on a lot of embedded boards.

      1. Unfortunately, the wacky world of ARM is a long, long, long, way from clean hardware abstraction.

        Some are weirder than others; but basically every vendor(and sometimes multiple product lines from a given vendor) is a port unto itself and the available quality varies widely.

        x86s can get wonky enough, and that’s effectively just two meaningful vendors building gear under the constraint that ‘if Windows won’t boot, we can’t sell it’. In ARM land, if you can vomit out a rickety BSP based on a kernel older than god and some probably-illegal GPL shenanigans, you can call that ‘support’.

        I’m hopeful that we’ll see some improvement in this area; Linus and company have certainly been doing their best to bludgeon some sanity into the process; but it’s still pretty ugly right now.

          1. Not confident in the “most are running Linux” statement. Consider all of the ARM based microcontrollers such as STM which actually make up the bulk of ARM sales, which are NOT running Linux. Many in that market segment not even able to run Linux.

          2. Yes, the trouble is that a great many of the ones ‘running linux’ aren’t running anything mainline, not even a pitifully antique flavor; and are stuck more or less forever on the vendor’s necrotic and relatively short-lived fork. Some vendors are better than others, and some devices eventually get brought into the fold; but quite a few only, and only ever will, ‘run linux’ in the very specific sense of the antique hackjob provided in the vendor BSP; and possibly some 3rd-party rebuilds based largely on what can be scavenged from that. The problem isn’t marketshare, the problem is that a ghastly percentage of the parts that ‘run linux’ have essentially zero support and little hope of timely vendor updates or mainlining.

    1. As you can see, the Odroid-C2 ships with a goodly-sized heatsink already attached. What little I’ve been on their IRC-channel I haven’t seen anyone say anything negative about the thermals, so I would assume they aren’t a problem.

      1. It’s fine without a fan if you’re not doing REALLY heavy workloads. At extreme levels, you may want a small fan, but even more so than the c1 before it for any normal use case it’s totally unnecessary.

  5. I’ve been eyeing the Odroid-C2 myself, but.. well, the problem is that all the graphics- and video-related stuff (3D, video-encoding, video-decoding, general X-acceleration) are in even worse shape than on the RPi. At least at the moment it has no H/W-accelerated encoding or decoding and no 3D under X at all.

    The RPi has poorer specs, but at least some bits of its GPU have been opened up and it’s more likely to have some longevity if you want to do anything graphics-related, but Mali is totally closed and fucked up and I don’t know if the video-engine on the S905 is any better.

    1. Yup, the problem with all these boards is in the graphics engine. Unless something changed recently, there is not a single board out there with open firmware that does not need a binary blob – with all its security issues.

      I believe there is enough demand for a free engine on a FPGA or similar device, capable of decent acceleration both for desktop and video purposes. It doesn’t have to be on par with commercial engines, frankly I don’t care if my cpu usage climbs at 80% instead of 15% when I’m watching a video, since I’m not doing anything else on the board, and for most of us it’s not a problem if desktop graphics acceleration isn’t lightning fast, provided it is usable for normal purposes. Once a proof of concept is created, if there is enough demand, somebody could grab it and make it into cheap silicon in some chinese fab.

      1. An FPGA which can do this is much more expensive than the dedicated hardware, when it additionally has much less performance, it is a very high price to pay just for open source drivers. I probably would not pay it.

        1. I’m definitely not an expert in FPGAs, my point is that they represent a way to develop and test a concept before making it into a real chip, which should be the final step. FPGAs can be developed by common people and their cat in their basement whilst silicon wafers need real fabs. If a design proves to be interesting, chances are that someone will turn it into real silicon. At least it could become a kickstarter.

  6. The C2 is not nearly as much of a winner versus the Pi3 as the C1 was versus the Pi2. The C1 had very similar advantages, plus it was three months ahead of the game, and had all the peripherals the Pi2 had.

    I’ve enjoyed my C1 over the last year, but had to retire it from its intended purpose because it’s stuck on the 3.10 LTS kernel and forever will be. It’s not commercially viable for Odroid to port an end-of-life product to a new kernel. The C2 starts with 3.14LTS; we’ll have to see if/when that gets updated. But even the oldest Pi1 can rock a 4.3 kernel.

    Even so, I might still have given the C2 a spin if the official UK distributor had been reasonable. £50=$70 for the C2 (including tax & shipping)? You’ve got to be kidding me.

    1. Yeah. I’m not sure what the deal is with that pricing. In the US Ameridroid charges at most a few dollars more than ordering direct from Hardkernel and the shipping costs appear to be the actual cost which is very reasonable. I have a hard time believing it costs that much to import into the UK but perhaps I am wrong being as I have never tried.

      1. Nothing wrong staying with an old kernel, until you encounter something that’s actually broke. The C1’s key job was to act as a backup target for my NAS. I recently switched the latter to btrfs, to find that the C1 fell over as a btrfs receive target. I was not too surprised: btrfs has seen a fair bit of development. A Pi is now doing the job happily if not quite as efficiently.

  7. Main problem with the C2 is that it currently ships with an old 3.14LTS kernel which is EOL in August this year. The Pi is currently already at 4.5 in development terms. HK and AMLogic are apparently going to start work on a 4.4LTS kernel in May – which is not a moment too soon.

    The issue with ODroid/Hardkernel stuff is that the level of support is orders of magnitude less than the Pi series. You may get more powerful hardware, but you have to solve a LOT more problems yourself.

    Fingers crossed the C2 doesn’t have any design faults (The original C1 had a few that meant the C1+ revision was required)

    1. Ordroid always SUCKED when comparing software side. What is worse that they do not care about upstream and don’t plan to merge support into upstream kernels.

      “going to start work on a 4.4LTS kernel in May” – and that will be their, yet another, FORK.

      I hate such forks… they usually get abandoned and you never get properly supported kernel.

      1. Even if they don’t make the effort to upstream the work themselves, basing it on 4.4 will mean that someone else should be able to do so with only a little work (unless they write truly awful code).

    2. They are already on revision 2 of the C2 PCB, which is a good thing in my book.

      2GB of RAM vs 1GB on the RPi3.
      DDR3 RAM vs DDR2 RAM on the RPi3
      10/100/1000Mb/s vs 10/100Mb/s on the RPi3
      2 real USB 2.0 ports, unlike the RPi3’s 1 (both have a 4 port hub connected to 1 usb port).
      eMMC+MicroSD vs RPi3’s MicroSD only.
      2GHz quad core A53 vs 1.2GHz quad core A53 on the RPi3.

      The only advantages I see to the RPi3 is bigger community, WiFi, slightly cheaper.

      I’m liking the direction of these new low power boards. I’ll probably wait until I see 4-8GB RAM on boards to replace my desktop with one.

      I’ve been slowly turning away from the whole telemetry OS and backdoor CPU’s of late, and I currently see an ARM desktop as one potential future. It still has blobs in the GPU’s, but it is the lesser of most evils.
      https://libreboot.org/faq/#intel
      https://libreboot.org/faq/#amd
      Builtin (potential) backdoors in CPU’s (a CPU inside the CPU with full access to everything your computer does, but you have no access to see the encrypted code it is running). You must “trust” US companies that it is all safe.

      1. The OpenRex schematics and other open source design files are available for free download.

        – Up to 4GB soldered DDR3-1066 RAM (533MHz)
        – NXP (Freescale) i.MX6 processor, up to 1.2GHz / 4 cores
        – SATA connector.
        – Networking — Gigabit Ethernet port
        – Arduino-compatible header
        – Raspberry Pi-compatible header
        – mini-PCIe slot with x1 PCIe and USB, plus a micro-SIM Card slot
        – accelerometer, compass, gyro; humidity and temperature sensors;

        http://www.imx6rex.com/open-rex/

        1. From an I/O standpoint, the iMX6 is wonderful — and beats the Pi hands down. I do development work on one basically every day. The video performance of it isn’t quite up to modern standards anymore these days though — and drivers for accelerated video are basically non-existant (and specs aren’t available without a separate NDA). I have both an RPi2 and a Cubox (i.MX6 at 1.2Ghz) and the RPi2 beats it handily for most video decoding tasks (especially for MPEG2 decode — e.g. 720p/1080i ATSC video). The other issue with the iMX6 is one of thermals — they basically run HOT, and I do mean HOT. When attempting to watch a movie on my Cubox, it would typically reach the maximum listed thermal rating for the non-commercial part code on the board within about a half hour and would then typically either freeze or reboot. I finally had to add a fan to it to keep it stable. So, the i.MX6 has some definite advantages (SATA, GbE, USB OTG, etc) for many purposes, but is unfortunately not up to the same level as the RPi2 when used as a generic media player with Kodi, etc. That said, on the totally custom i.MX6 board designs we use at work having decent NAND on either GPMI or especially WEIM is unbelievably faster than the crappy SDIO interface that the RPi2 is stuck with. It is also nice to be able to have it all formatted with something MUCH more reliable like UBI.

        2. My only issue with the i.MX6 is as it states in Errata ERR004512 that the maximum performance “is limited to 470 Mbps (total for Tx and Rx)”, which is not really gigabit in my book. It is a bit like claiming that the RPi has 4 USB 2.0 ports, when it only has one real USB 2.0 port connected to a hub chip (Microchip LAN9514 or LAN9512 on the older boards) with inbuilt 10/100 NIC.

          REF: i.MX6DQ Errata document – https://cache.freescale.com/files/32bit/doc/errata/IMX6DQCE.pdf?fpsp=1&WT_TYPE=Errata&WT_VENDOR=FREESCALE&WT_FILE_FORMAT=pdf&WT_ASSET=Documentation

          1. Oh and even with saying that I’ll still probably support the openrex project, because I’d like an ARM board that has the potential to run OpenBSD.

        1. I’m afraid you’re terribly misinformed. TrustZone on ARM plays a completely different role to the various Intel embedded controllers.

          TrustZone is an (entirely optional) architecturally separated enclave for processing secret data, used for things like DRM and payment security. It’s not security-through-obscurity either – the OP-TEE code is even on Github: https://github.com/OP-TEE
          Importantly, it’s under the control of untrusted software – the security doesn’t rely on preventing untrusted code from accessing the TEE APIs, because it is enforced on a hardware level.

    3. The old C1 was quite usable however. I have several and they have performed flawlessly for me. I have just recently ordered several of the new C2’s and am looking forward to good performance from those– fingers crossed of course ;-)

      1. I have a different experience. I just have one Odroid C1, bought it to build media center, but every other update of the Ubuntu seemed to break it (Kernel Panic, no boot etc.). I put it in a drawer too angry to look at it after hours of lost time. Stayed there for months. Yesterday I dusted it off and put Android on it. I hope it will be a better experience than with Ubuntu (so far so good and even Kodi performance seems much better on Android than on my previous Ubuntu installation).

    4. More and more embedded processor manufacturers learn the hard way that there need to maintain chip specific Linux driver directly into the Linux mainline. Too many are still trying to make BSP/SDK with a Linux distribution that do just a few easy examples based on a random picked kernel that was modified on purpose for months and thus outdated. Linux developers already know how to make a distribution, often better than the SDK, and to share it to the community. Manufacturers really need to only focus on the mainline drivers support (and boot loader sometimes).

    5. Agreed the support on the Odroids needs to be improved, Good magazine, good forum, reasonable attitude but the Kernel and dirver support is poor as well as no one seems wants to play well with getting the stuff up stream in to the normal release Kernel

      1. The way I understand it, all kernel dev at hardkernel is handled by two people. One guy in Korea does work on the kernel provided by the SoC manufacturer for use with Android, then mdrjr does all the Linux distro kernel work from there. At least, that was how it was being done a couple of years ago.

        Hardkernel just doesn’t have the budget to hire droves of Linux kernel devs, nor is the community large enough to have already solved every possible software/driver need. Their small community is, however, very attentive, helpful, and dedicated, and several driver issues were solved by community members for their Exynos-based boards.

        Actually, I believe mdrjr was actually hired by Hardkernel right out of the community. The ODROID magazine is also a community effort.

        So, while there are drawbacks to choosing an ODROID board, you ARE getting a great community with it. They just aren’t as big as RPi’s.

    6. > the level of support is orders of magnitude less than the Pi series

      Buh? No. They have a forum and the HardKernel people participate and will answer question. For RasPi, pretty much *everything* is just some esoteric hack that someone in the community did, and there is a blog post floating around out there somewhere about it.

  8. The bigger problem with ODroid is that they asking around 60 Euros/65 dollars for it in Europe. The options and namely the eMMC is charged to the double of price you can buy it in USA or Korea. If you try to import it, you are in for huge taxes. On top of that, the shipping charges either in any of the stores, korea, europe, or USA are damn insane. So buy the time you buy an oDroid in Europe, one way or other, it will cost you around 80 euros without any options…put an eMMC there, and you are paying to the tune of 150-200 Euros for the same kind of thing that will cost you in USA around 80-100 dollars. Utterly insane. So no Odroid for me in Europe. They should *really* talk with the banana pi guys and find out how they send stuff to Europe…

    1. I was going to blame socialism (and the necessary high import taxes to keep socialism viable) for the price problems in Europe, but if the Banana Pi guys are getting them there inexpensively, maybe there’s a loophole. I’m not trying to start a politics/economics flame war, so don’t even…

      1. The Banana Pi line is of lesser quality and they also don’t pay licensing fees that they’re legally required to. They slip through the cracks because a ton of ARM hardware providers do the same thing. Playing by the rules actually puts companies at a disadvantage, which is the opposite of what is supposed to happen.

  9. Price and power is not the let down on these. It’s the community around them. I dont care if the board is 900ghz 16 cores with 10,000 IO as well as a cellphone transmitter on it. If it does not have the huge community around it making it effortless to use it will be a giant failure.

    Look at the Olimex boards. They are GREAT hardware at great prices but almost nobody on the planet knows about them and they sell units in the Hundreds not millions. And when you ask specific questions about hardware modifications all you get is “study the schematic” and RTFM NOOB! which makes people walk away from it rapidly.

    So for this board to survive, they need to pour 5X the effort into making it into creating the community. The engineers need to answer even stupid questions and hand hold people themselves and spend hours daily on the forums. they need to hire people to write clear and concise guides, etc….

    1. I’ve had the developers and helpful uses both answer my often idiotic questions when I got my first C1. I found the community, while smaller than the Pi’s community (for reasons often totally outside Hardkernels control) they were absolutely willing to help to the best of their ability. This included the manufacturer outlining hardware modifications to get CEC working (which was never an advertised feature of the board)

  10. As far as external storage is concerned on the C2. Since it has a gigE port on it… iSCSI would be perfect..

    Now… if could power it using PoE or even PoE+.. That would give it enormous potential!

  11. The other thing that makes me hesitate when it comes to odroid is how quickly they depricate and discontinue boards (and stop providing updates for the handful of drivers, etc… Their updater facility gathers from their website. Since the drivers are not in the main line kernel and they count on plugging opaque binary firmware blobs into exact kernel and module versions you get locked in to an ancient kernel, or you can buy a new board and get locked in with it to a newer kernel.

    This effectively comes down to “Is it a user-facing general purpose computer or is it a beefy embedded controller ?” The first needs to stay current to be useful, the second is as likely as not meant to live its life headless and airgapped performing some vital task where the risk of breakage from .so rot and interface compatibility outweighs the downside of old (but mostly stable) software.

    1. This is the thing people need to remember. The RASPI seems to me to be lacking in many ways on the hardware front, particularly in the power supply, and the methods used to ‘bootstrap’ vital interfaces like ethernet onto the board. Its kind of shameful that the pi is marketed to some extent as an industrial computer, as software stability and solid interfaces are usually at the top of the list of requirements for embedded industrial computing platforms. Odroid could win hands down if they upped their game on the software and community support front. As far as price in the EU, talk to your EU regulators why they cost so much! The Pi is made in the EU, so it is not subject to the taxes and import duties like the Odroid.

      I have been tossing around which one I should get for a network controllable rover I want to build, and I am honestly leaning toward the Odroid.

      As far as deprecating boards, RPI is not much better! I don’t think there are any of their boards that have been marketed for more than a year at most. The form factor of most of the boards leaves a lot to be desired too, particularly the connector arrangement. The only thing they have going for them is that the software they package with them is slightly better supported, so the growing pains of moving to a new platform are mitigated to some degree.

      In any case, the availability of such powerful computing platforms for such a low cost is changing the face of technology, and it looks like people who enjoy this stuff have nothing to fear but things getting better and more accessible!

  12. I just bought a couple of odroid-c0’s and I’m very happy with them .Having said that, from now on I will only purchase SBCs with proper Linux Kernel maintenance and support with a decent mainlining effort. RPI3 and Dragonboard 410c are two excellent candidates. (The RPi3 costs less than half of the Dragonboard 410c and has almost the same specs though.) I’ll consider ordering C2s only when I see a significant effort being put into mainlining the Kernel.

    Having said that I believe that the quality of odroid boards hardware wise is superior to even the Raspberry Pi 3 boards. and the community for these boards is huge. Not quite as large as the RPi community but is still large enough to be able to answer many questions for new users.

    HardKernel even releases a free monthly magazine called ‘odroid magazine’ very similar to the magPi http://magazine.odroid.com/

    And don’t forget Arch Support. Almost all odroid boards have excellent Arch support https://archlinuxarm.org/ …as do the RPi 2 & 3 boards as well.

  13. The first thing I think of learning that the next Pi includes wireless is “do those features continue to use power in any way after I turn them off?”. Second I start wishing they had used that space for other features such as mic & line-in audio, SATA, USB-3, jsut about anything else!!!

    I mostly see Pis as being usefull as mini-desktops and personal servers. For that I will use Ethernet thank you.

    Sure.. Pis were made for education. I don’t know any kids that are without access to a computer. I don’t really understand the point of a Pi for that. If I did know any such children then I suspect they don’t have a TV new enough to do HDMI either. If they do… then there parents have some pretty bad priorities!!! I see $10 VGA monitors in the local thrift shops all the time. An ‘education’ computer really should be able to do VGA! Composite is just so low-res. That should be an absolute last effort sort of thing to use.

    Ok.. so Pis seem to have some secondary goal of being IoT devices. They have lots of GPIOs.. that’s nice. Are you going to run a real time OS on your Pi? I’m guessing no. The way most people use them for IoT.. Linux and Python it would probably be better to just hang an Arduino off of it and just treat the Pi as a computer controlling an Arduino. That way at least.. when Pin A needs to turn on NOW and Pin B needs to come on after Pin A.. you actually KNOW that Pin A WILL turn on NOW and that Pin B will ALWAYS come second. And stuff like that…

    Portable devices… Ok.. MAYBE somebody will want to make a handheld device out of a Pi. No hobbyist SBC I have seen is ideal for that though. How about a RasPi portable edition. Get rid of those giant connectors.. they make it too fat! I’d replace the USB and Ethernet jacks with pin headers and the hdmi with a micro-hdmi. Actually… no soldered in headers… just holes for them, like a Pi Zero but not so cut down. Let the user solder them if IF they want them.

    For this version of the Pi built in wifi and bluetooth finally do make some sense. But.. now you need an audio input more than ever… at least a mic jack if not line-in. Line in should be almost free if you are already doing a Mic jack though so how about both?. Make all those audio jacks header holes too by the way… Who wants to have to plug a speaker and mic into their handheld device? A plug socket, if it exists at all should only be there as a way to add an optional accessory. This way the user can easily just solder on a speaker and mic. (this was a major failing IMHO of all the Arduino cellular modem shields I have seen).

    Or.. maybe.. if you can fit it.. actually put a tiny speaker and mic on the board! Please make them disableable through some jumper though. The jumper header pin can also act as an access point to connect external speaker/mics. Also.. make sure the speaker and Mic are on OPPOSITE ends of the board. Come on.. think cellphone here.

    One more feature is necessary for this ‘portable device’ edition of the Pi. It should have a battery charge controller! I really don’t think anything more is necessary for explaining that feature. Seems pretty obvious doesn’t it?

    If adding these things means something has to go then fine… do away with the ethernet first.. that’s the easiest to replace via USB. Then lose the HDMI next if absolutely necessary. Portable devices are going to use the LCD connector and WiFi anyway. But… convergence is Awesome! If my handheld, portable device can dock and become a computer then that is way better. So.. only eliminate those things if absolutely necessary!

    1. I don’t know what area you live in, but for me, those $10 vga monitors are usually about $30 to $50 (thrift shops in my town have weird prices on a lot of stuff), but almost every single tv and computer in the house either came from someone’s trash or donation, and I’ve fixed or upgraded them with spare or inexpensive parts. So I’ve got some good tech in the house, but I assure you my priorities are in order. I’m guessing any parent who would get a Pi as a learning computer for their kid would have some similar interests and/or abilities.

      1. Well.. NE Ohio. I know one thrift shop that pretty much always has at least a couple of them for $10 and several others that occasionally do. On the other hand.. Do those TVs you get cheap/free come with HDMI? The only thrift shop TVs I ever see are old CRT ones. I don’t know anybody that is giving away TVs. Even getting so much as a composite input without also buying a modulator would be a lucky find! I guess you and I live in opposite lands!

        So.. about your response to my statement about priorities. Are you actually unable to afford a ‘regular’ desktop for a kid to learn on? If not, and yet you have a house full of TVs sucking up electricity and maybe even using cable or subscription satellite service then maybe I was talking about you. I’m guessing that is not the case. What are you even using to get on HaD?

        That was my impression anyway of what the original goal of Raspberry Pi was. Get some form of computer into the hands of kids in families that couldn’t afford a computer. Except.. here in the first-world anyway who is that? At least in my neck of the woods old office computers are so plentiful that it is very difficult to even give one away. One might have to buy a monitor but that is about it. So.. if someone can’t afford that then I question if they are more in a position of needing a computer or one of needing food, shelter and medicine!

        Yes.. I realize this is my first-world intrepretation. But is a RasPi really any good for people in third-world countries? Do they have plentiful HDMI televisions and easy access to all the hookup cables, SD card and other accessories needed to make a RasPi work? I would expect OLPC makes more sense there but I admit, I have no real experience in that area.

    2. For portable devices, the C0 (C-zero) comes with most ports unpopulated and they sell a connector kit for a couple bucks that has all the connectors you might need so you can make it exactly like you want it. Plus it has a built-in LiPo circuit so it can be run off a battery, and the battery can be recharged without disconnecting it from the C0. I’ve helped develop a C0 walking robot with a touchscreen “face” (you can see the previous version based off the C1 here: https://youtu.be/A7kuNEi02gQ) and a minimalist quadcopter is in the works.

      1. I assume that you were responding to what I wrote about a ‘handheld device edition’ Pi. The more I think about it.. what is really needed for that is just a pre-built device, not a dev board. The problem is.. for a true, handheld device space requirements are so constrained. You really want everything on the mainboard. Every plugin adds not only it’s own width but the width of the connectors that join them. The C0 is close but it is missing Audio and LVDS. Plus.. since it is missing Audio it would be especially bad because it only has 2 USB ports. You would want 1 port open for external accessories. Another port is for the touch screen. Since it is missing Audio… you need a third port for that. So now you need a USB hub too!

        Add the battery and LCD and what you have is a brick!

        Once you add all that onto the mainboard.. and an LCD perfectly sized and positioned to plug into it you practically are selling a completed device. So.. how about this… an open source tablet with either a built in Arduino, pin headers accessible but female rather than male to make it safer. Or.. instead of the Arduino a built in Bus Pirate. Might as well make the external USB port full-size. It isn’t going to be as sexy as a new cellphone but come on, a single USB port isn’t THAT thick!

        Maybe a third modle could have a fatter case with an internal USB port. I’d buy that. pull the board out of a cellular modem stick (to make it thinner) and plug that in there. Then it can be my cellphone!

        Come on, somebody here has to be able to design and build one of these! I’m still too much of a PCB design noob to tackle somethign like this. How about a kickstarter?

  14. As someone who uses these single boats computers for fun and to build relatively low bandwidth things anyway, I’d MUCH rather support the Pi Foundation in the UK and their mission than a company in Korea I’ve never heard of. The Pi foundation has so many tutorials and things to help people learn, for me, it’s not all about the specs.

      1. The “Arduino” of the SBC world

        So.. questionable technical merits but…
        vastly superior community support… especially support for newbies…

        Yup. I’d say that describes it pretty well. But.. isn’t that pretty much what they always wanted to be? I’m not sure I see the problem here.

        From the perspective of being intended as an educational device I think that newbie friendly community is exactly what they need. From a marketing perspective.. at least in the low-priced hobbyist market I’m pretty sure Arduinos (and compatible stuff) way outsells everything else.

  15. That looks nice, the GBit either net will be very good addition. Be nice to know if it would make a good NAS. Reading the specs is not good enough, you need to try it to be sure. I tried a RPi, switched to a cheap second hand atom system instead.

    If you like the Pi or not, they shock up the SBC market. I’ve been working with small ARM prototype systems on and off for years, normally with preproduction chips. Before the RPi the asking price was excessive, hundreds if not thousands of dollars. Now, the price of these dev kits have also come down in price because people would simply by a RPi instead. And if you want people to dev for your platform, putting a big price on the dev kit is a bad way to go about it. Not all the chip companies have caught up, some still asking daft prices.

    Disclaimer, I have 10 RPis including two Zeros. I need help. ;) LoL

  16. By the time you pay for the courier and their brokerage charges if you are outside of Korea/US, one of those Intel ITX boards (e.g. quad core N3150 6W) is starting to look attractive. You’ll get better Linux support, 4K Intel HD video, 6Gbps SATA, memory slots, Gigabit Ethernet, PCIe x1, USB 3.0.

    1. Pretty much a FAIL if it doesn’t support AMD and Intel chips. :P Linux and most software would just work fine without that.

      Most end users would just want something that works and not worry about the “free” part of it. Nothing is really completely open source and live up to the ideology as the chips, the technology libraries for the foundry are still closed source.

      1. There are some open designs like the OpenSPARC, which since it’s release has had it’s designed radiation hardened for use in space. But it is hardly a great processor by today’s standards, it was opened a long time ago..

  17. i like it, id like it more if it had wifi built in. the new pi pretty much frees up 2 ports which would generally be used for a wifi dongle and a bluetooth dongle. i tried a 2 in 1 dongle but go figure no pi driver was available. but with the pi 3 its there and immediately supported by their official operating systems. its not just the ports but power those dongles suck as well, suddenly the on board psu is sufficient to run everything you need without needing a beefy external buck regulator.

    of course im mostly interested in wireless applications. if i was looking for a miniature server or media center or anything i could just plug in and forget about, i would be looking at this thing.

    1. If power doesn’t matter and wireless does, then the pi3 is certainly superior. Slightly cheaper out of the gate, and saves $10 on dongles (assuming the built in wireless adapters have reasonable antennas).

      Personally, though, even though every place I need one will use Wi-Fi, and/or Bluetooth, I’m happy to pay the extra couple dollars to add them. I can’t add 800mhz and 1gb more ram to a pi3 :(

  18. The cheapest that you can get a C2 from Ameridroid is $48.70 with shipping. Considering that it costs $13.70 more than the Pi 3, it’s not really an apples to apples comparison in any stretch of the imagination.

    1. And yet I’ve found that Pi’s never end up being sold for their list of 35 either. Maybe in the US, but certainly not elsewhere. The c2 lists at $40, after all, $5 more than the pi. I bet I could fill a truck with all the people selling Pi3’s for ~$40

    2. $48.70 is the cost with shipping Priority. It is $41.95 without shipping. It isn’t fair to compare the cost with shipping against a $35 Pi without shipping. I can’t get a Pi shipped to my door for $35. And certainly not using Priority Mail. I’d be lucky to get a Pi for $40. Using Priority/Ground the cheapest is probably going to be ~$42.

  19. I’ve been fairly happy with my Odroid C1’s, they were certainly far more performance than the PI2 at the time. I have 3 of them running as headless home servers, and apart from some niggles with SD card compatability they fly along, sipping the tiniest amounts of electricity.

    Seems like they are different markets though, on-board Bluetooth and Wifi in the PI3 will be indispensable in some projects.

  20. So.. a lot of people talk about the community and support behind hobbyist SBCs. I just bought my first.. a Banana Pi and haven’t even powered it up yet. I chose the yellow fruit because of the SATA port.. my first task is a low power file and LAMP server.

    Anyway.. maybe it will become aparent once I actually try to use the thing. What is this support stuff all about? Isn’t it just Linux? Ok, so the install is a little different, ‘burn’ an image to an SD card rather than run an install CD. Once you get past that what makes these SBCs so special that they need different support from say… running Debian on an x86 desktop? How different is it to use one than another? Are they really all special and unique snow flakes?

    Not having personally used one yet.. all I see is a little computer running Linux.

  21. Installing Linux on x86 is easy. x86 cores are fabricated by two companies and are identical. All other supporting hardware and subsystems are also similar with very minor difference between AMD and Intel x86 devices. ARM is different. Sure the cores are identical from one SOC to the other…say A53s on an RPi3 (BCM2837) vs a DragonBoard 410c (Snapdragon 410). But all of the supported subsystems (Ethernet, USB, GPIO, GPU, I2C e.t.c) can be profoundly different from one SOC to another. This means that you simply can’t take a Linux Kernel or distribution for the RPI3 and just use it for the Dragonboard.

    This is why mainlining ARM SOCs is absolutely essential but with so many varied number of SOCs with different subsystems requiring different drivers this is a lot easier said than done. In fact it can be quite a shit show actually. Ideally every ARM SOC manufacturer should have one Linux tree that includes kernels with drivers for all of its SOCs and should put some effort into mainlining that Kernel tree i.e. try to get it accepted into the mainline (Linus’s) Kernel.

    But this is so much easier said than done and most of these companies especially Chinese companies that might not appreciate the value of mainlining. Furthermore they develop separate Linux Kernels per SOC, and refuse to put the time to maintain them after the SOC becomes older. Many of them only need the Kernel for android anyways so the whole Linux Kernel GPL and open source philosophy thing is very much under-appreciated and in some cases completely misunderstood. This is why I refuse to buy any SOCs from companies that don’t have a well maintained Linux tree and that do not try to mainline their kernels e.g. Allwinner & AmLogic.

    1. I keep hearing that AllWinner and AmLogic are planning to change this 3 months or 6 months from now for the last two years now but rarely see any efforts from the companies themselves. Instead I only see efforts from the open-source community e.g. linux-sunxi which is amazing but I feel that this should be coming from the companies. Also this sort of development can be slow when being managed solely by the community with little to no backing from the SOC manufacturers themselves.. So I’ll wait and see. I’m hoping that significant mainlining efforts will be made via the CHIP and PINE64 kickstarter projects….both AllWinner based. And hopefully AmLogic will follow suit.

      Another problem is the longevity of the SOC/boards. For example TI’s pandaboard has pretty decent Mainline kernel support but isn’t being manufactured anymore!!!! This is why I like the Raspberry Pi Boards. I know that five years from now, I’ll be able to get a linux distro that runs on them and will be able to access the latest kernel for the board when i need it.

      1. It seems to come down to either having longevity or being cutting edge. The Pi’s have a very long support cycle but they are not very cutting edge. The Odroid’s are the opposite with shorter support cycles but are also one of the most cutting edge sbc’s available. Hardkernel typically makes fairly significant changes between boards.

        The Beaglebone would be another long term support board. You can still get their latest images for the original Beaglebone. But they are definitely not very cutting edge anymore. Kinda sad to see my favorite board falling further and further behind because of dated hardware.

        1. I love Hardkernel and the Odroid board’s. The boards that were based on the Samsung SOCs e.g. odroid-U3 and odroid-XU3 are excellent and seem to have much more longevity than the ones that use the Amlogic cores. Sure Amlogic cores are cheaper but don’t have the longevity nor the support to justify using them for more than a year or two. They’re basically cheap ‘throw away’ SOCs to be used in the current latest model of Android tablets and TV set-top boxes and they keep getting switched every year and some instances every six months.

          To an extent this is the case with almost all ARM cores. But AllWinner & AmLogic SOC manufacturers suffer from this more than the others. Furthermore we’re getting to a point where even the newer model SOCs offer only incrementally better performance than the older SOCs. For example if you compare the performance difference between the new Nexus 5X (Snapdragon 808) and 2-3 year older Nexus 5 phones (Snapdragon 800)….there’s really not that much difference.

          Perhaps its time to build a Reasonably performant SOC; say something based on Cortex-A53/A72, 2GB+ of RAM built-in WiFi/BLE, Gigabit Ethernet, USB3.0, built-in EMMC, with proper long term Linux support and proper GPU support for a decent price (under $50) and stick with it for a period of 5 years.

          What many of these ARM SOCs lack especially in the SBC (single board computer) arena is a degree of standardization. No one says that everything has to be dictated by one or two companies like in the x86 world. But what we have here is indeed quite chaotic. And this is the true void that the Raspberry Pi3 has inadvertently filled. Sure the original motivation of the Raspberry Pi foundation is encouraging youth to get into the computing field, but the true reason that the Foundation met a high degree of success is that they brought a standardized easy-to-use ‘platform’ to the SBC world. Good on them.

  22. Thou Odroid C2 beats RP3 on most of its hardware, it is the warranty, community and the compatibility with 3d party products that made me buy yet another RP, this time a RPI3 for my audio and home-automation projects. However, I ordered an Odroid C2 from Germany to see if it is useful. This time I will use it as a development platform for package deployment on production RPI3s. It is the time of the compilation of s-w, drivers as well as running Eclipse that made me considering trying a C2. I want better audio boards for C2, such as the IQAudio stuff.

  23. Well, I have both right now. I’ve got the C2 for 2 weeks now, and I bought the RPi3 today. What I can say for now is that the C2 is fast, it gets the highest MIPS per dollar ratio among all the boards I’ve tested to date. And since I’m interested in high performance networking and CPUs for network stress testing and build farms, it seems unbeatable for the price for now. Also, it worked out of the box and remains barely warm even while building kernels with the 4 cores at 100%.

    The RPi3 I bought today simply doesn’t boot from the available raspbian images from the official site :-( I’m amazed that a product apparently targetting beginners panics at boot in such a cryptic way for beginners (unable to mount root from unknown-179:2). Also, the pointers I’m seeing in the kernel panic are 32-bit, which tells me that they didn’t even port their OS to 64-bit yet. The Noobs loader that’s supposed to do everything by itself didn’t detect the wired network, failed in the middle of the install over WiFi and then failed to detect WiFi after this… The product clearly looks rushed out with very little qualification for now. I’d advise to wait for a bit of settling down.

    I think the RPi3 can become a good product overall but for now it looks more like a quick and unfinished upgrade to the RPi2 unfortunately. Let’s wait a bit to give it a second chance in a few months…

    1. Your RPi3 experience is very untypical indeed. I’m wondering what’s up, it may be a dodgy power supply or even faulty hardware – it happens. You are right that no 64-bit kernel is yet available for the Rpi3. Odroid makes great hardware, but you can’t beat the Rpi on community, software, and longevity.

      1. I agree, and that was the reason of my choice, it’s for offering. BTW, I found the root cause of the problem : the two different microSD I tried are damaged (one is a fake 8GB limited to 512 MB and only supports the Noobs boot loader, the other one seems defective after ~2G). That allowed me to boot a lite image on the <2GB one confirming the card works. I found the issue by slowing down the messages on screen to have the time to take a capture to see the first error. I couldn't indeed properly enable a serial console yet. So for now the RPi3 is not to blame apparently.

  24. I was deciding between the RPI3 or Odroid C2 for my first SBC. Price wise they come out the same if not in favor of the Odroid C2 since it includes heat sink and uses common 5V 2A power which I have many lying around whereas the RPI3 runs hot requiring purchase of additional heat sink and uncommon 5V 2.5A power. Specs wise the Odroid C2 is far ahead with 2GB DRAM vs 1GB, faster performance, line rate gigabit ethernet, eMMC 5.0 storage option, infrared receiver, has 5 USB ports if using DC power and support both Linux and Android. After owning the Odroid C2 for a little over a week I’m very happy with the purchase and being able to choose between Odroibian Debian 32bit Linux for server and desktop replacement, Android 5.1.1 for media consumption and access to all the Google Play apps and Kali Linux security distribution. Was tempted to pick up a RPI3 for $30 ($35-$5 off coupon) from local Micro Center but I think I’m going to get another Odroid C2 or perhaps an even faster XU5 when it comes out.

    1. The Pi3 works absolutely fine with one of your spare 2A adapters unless you’re planning to festoon the USB ports with power-hungry gadgets – in which case the C2 will struggle on 2A, too. The board itself doesn’t draw more than 800mA or so when running at full tilt. The C2 has almost identical power consumption.

      There’s no question that the C2 has the better specs by some margin. Though I hope that you’re running a 64-bit OS on it! For me, I got myself a Pi3 to play with. Why? My C1 (which never worked completely without issues, so I wasn’t surprised by the C1+ iteration) is gathering dust because its ancient kernel choked on what I wanted to do with it: offsite btrfs snapshot backups. The job is now done by a Pi2. The maturity and support of the Pi platform can’t be beaten right now.

      Small but perfectly adequate copper heatsinks for the Pi3 are less than $1 apiece from Aliexpress.

      1. I picked up a RPi3 for cheap at Micro Center. Having used it for a few weeks it’s not only slow but also prone to thermal throttling with normal workload such as playing a 360p windowed video and very picky about power. It only wants to be fed by a Samsung 5V 2A adapter with very short about 1.5ft thick USB cable otherwise it’ll flash the rainbow square voltage warning and even then it locks up when running xhpl benchmark. It’s rubbish.

  25. A few other thinks to consider –
    A full schematic is available on the C2 (and other Odroid boards) including the FULL schematics for not only the current version, as well as the two previous releases.

    This is actually significant in TWO ways that make the Odroid folks look very good, and make the Pi Foundation look VERY BAD by comparison.

    FIRST, the Odroid folks have released ALL their schematics, where the ‘open source’ Pi Foundation still has not even released the schematics for the Pi 2 (much less the Pi 3).

    … and SECOND, by releasing multiple version schematics, the Odroid folks are at least trying to practice responsible version control, with responsible disclosure practices – the Pi Foundation on the other hand, will never even admidt that hardware bug fixes are being done, or that changes are upcomming untill they “sell through” all their defeto-duds like they did with version 1 of the 7 inch touch screen LCD (the version with the DEFECTIVE BACKLIGHT PCM CONTROLER). Folks were desperately asking for ANY info on these LCD’s for weeks on end, PLEADING for info on how to make the backlight control work correctly, BEGGING — While the Pi Foundation kept mum, while knowing full well THAT THE FIRST RUN LCD TOUCH SCREENS WERE JUST FLAT OUT DEFECTIVE. I believe they did this crassly so folks couldn’t dodge the bad units in favor of the working version 2 units, and force them to eat the cost of their mistake in the form of returned stock from dealers. As soon as all the Defecto-Dud displays had ‘sold-trough’ into the retail channel (and were beyond the possibility of return) then ‘magically’ the correct information became available. (that version 1 will NEVER work, only version 2)

    The Pi Foundation fan boyz point to the iterative hardware releases of the Odroid series like it’s a of sign of an immature company that ‘just can’t get it right the first time’ while ignoring numerous similar Pi Foundation blunders that got swept under the rug.

    40 years of experience tells me that glitches always happen, but it ALSO tells me that dealing with them is a lot worse if you are not dealing with paranoid dishonest people, that won’t even release their boards schematic so you can help troubleshoot, and who give dishonest self-serving answers to purely technical questions.

    My first experience with this was when a Pi Foundation spokesmen said that there were “no plans for an upgraded Pi version” (or something to that effect) only weeks before the significantly improved Pi 4 Port USB model was released (causing me to order and get stuck with already obsolete hardware, where, if an honest answer had been given, I would have been happy to wait a week or two for the newer model). Yes, I do understand the commercial pressures that lead manufactures to dissemble like this – that doesn’t mean I have to like it.

    Then theres the whole issue of making you pay for 20 year plus expired patents for MPEG2. Yes there are one or two remaining patents in the “patent pool” run by the MPEG LA Nazis – BUT THEY AREN’T “ESSENTIAL” FOR SIMPLE MPEG2 VIDEO PLAYBACK. This is obvious when you think about it DVD’s were released more than 20 years ago using MPEG program streams with all the bells and whistles needed at the time for DVD playback. MPEG2 Video, AC3 Audio, DVD DRM, Subtitles – EVERYTHING. Since all these things have been in commercial use for more than 20 years now – THEY CAN’T POSSIBLY BE STILL LOCKED DOWN BY PATENTS. The remaining patents are all for esoteric addon crap that we don’t need, but that MPEG LA desperately wants to sell – and the Pi Foundation are too big of WIMPS to call them on it!!!

    So, after a lot of initial enthusiasm for the Raspberry Pi, the combination of the recent lack of schematics and other critical technical information, and even worse lack of honest answers, and other sundry annoyances, has me ready to call it quits with the “Pi Foundation” and I am happy that an alternative showed up that looks so promising.

    The thing that made the Pi work was community support. I never thought the Broadcom hardware was very impressive – slow, glitchy, and absolutely LOUSY power management (no micro-power sleep mode whatsoever??? wuzup wih dat???)

    Odroid at least seems to be trying to run a professional organization and to be responsive to their users, and considering their limited resources, I think they are doing an amazing job. If the Odroid series continues to improve at it’s present rate, and if it eventually gets even one tenth the community support the Pi does, then it will end up being ten times the performer that the Pi ever was.

Leave a Reply

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

WordPress.com Logo

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

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s