Review: The Asus Tinker Board (Updated)

In the years since the launch of the original Raspberry Pi we have seen the little British ARM-based board become one of the more popular single board computers in the hobbyist, maker, and hacker communities. It has retained that position despite the best efforts of other manufacturers, and we have seen a succession of competitor boards directly copying it by imitating its form factor. None of them have made a significant dent in the sales figures enjoyed by the Pi, yet they continue to appear on a regular basis.

We recently brought you news of the latest challenger in this arena, in the form of the Asus Tinker Board. This is a board that has made us sit up and take notice because unlike previous players this time we have a product from a giant of the industry. Most of us are likely to own at least one Asus product, indeed there is a good chance that you might be reading this on an Asus computer or monitor. Asus have made some very high quality hardware in their time, so perhaps this product will inherit some of that heritage. Thus it was with a sense of expectation that we ordered one of the first batch of Tinker Boards, and waited eagerly for the postman.

Update:

A member of the Asus Marketing team read this review and contacted Hackaday with some updated information. According to our discussion, the Tinker Board has not officially launched. This explains a lot about the current state of the Tinker Board. As Jenny mentions in her review below, the software support for the board is not yet in place, and as comments on this review have mentioned, you can’t source it in the US and most other markets. An internal slide deck was leaked on SlideShare shortly after CES (which explains our earlier coverage), followed by one retailer in the UK market selling the boards ahead of Asus’ launch date (which is how we got our hands on this unit).

Asus tells us that they are aiming for an end of February launch date, perhaps as soon as the 26th for the United States, UK, and Taiwan. Other markets might have some variation, all of this contingent on agreements with and getting stock to regional distributors. With the launch will come the final OS Distribution (TinkerOS based on Debian), schematics, mechanical block diagrams, etc. Asus tells Hackaday it is a top priority to deliver hardware video acceleration for the Rockchip on the Tinker Board. The Board Support Package which hooks the feature into Linux is not yet finished but will come either on launch day or soon after. This is the end of the update, please enjoy Jenny List’s full review below.

First Impressions

Just what you want the postman to deliver!
Just what you want the postman to deliver!

The packaging is as professional as you’d expect from such a large manufacturer, a small cardboard box with the Tinker Board logo on top. Open it up, and the board sits encased in an anti-static bag on a cardboard tray. Under the tray is a flimsy instruction leaflet, and a stick-on heatsink for the Tinker Board processor.

Unpacking the board it is immediately obvious that it is a faithful physical clone of the Raspberry Pi form factor. Unlike some Pi clones such as the Odroid boards it has micro SD card, camera, and LCD connectors in exactly the same places as those on the PI, and replicates all the Pi’s external connections. It’s a safe bet that this will fit inside the majority of Raspberry Pi cases. It does manage to fit in a little extra though, there is a detachable antenna connector for the onbard wireless chipset, and hidden between the LAN and USB connectors are a pair of unpopulated holes for PWM and SPD/IF audio.

The Tinker Board in all its glory.
The Tinker Board in all its glory.

On the board itself, aside from the Rockchip RK3288 system-on-chip, there is a Rockchip RK808 power management/clock chip, Realtek Ethernet and audio codec chips, and a Genesys USB hub chip. There is also a wireless module with that detachable antenna, handling both 802.11 and Bluetooth. On the underside are the RAM chips, and the SD card socket.

The aesthetic design is pretty noticeable, with the board featuring patterns and labels for each connector. A particularly nice touch comes with the GPIO header, its plastic is colour-coded to identify the different types of pin.

Getting started with the Tinker Board is a very similar process to doing so on the Raspberry Pi. Connect a suitably powerful micro USB charger to its power port, add an HDMI monitor and a USB keyboard and mouse combo, and you can boot an OS from a micro SD card.

A Few Potholes In The Road

At this point though, we encounter a snag. A Raspberry Pi owner is used to visiting the website, clicking on the download link, selecting an operating system image, and following the comprehensive instructions, but Asus Tinker Board users get the following instructions in their leaflet:

Download the OS image from the Tinker Board website and burn it into the micro SD card using a third-party ISO software, such as Win32DiskImager.

This all sounds extremely promising, but for one unfortunate omission. There is no Tinker Board website. That’s right, a huge multinational hardware company has produced a high quality product, and neglected to create the online component that goes with it.

For those prepared to search, there is now a section of the Asus driver download site on which you can find some downloadable resources including a pair of OS images. This was nowhere to be seen for a while after the product was available for sale, but appeared before packages started landing on doormats. Presumably someone at Asus was alerted to the dismal prospect of being panned for providing a product with no software, and that was the best they could do. We’re guessing most Hackaday readers will find no problems installing a downloaded image onto an SD card, which is handy because Asus haven’t provided any instructions. It’s probably fortunate then that the Raspberry Pi Foundation’s instructions will also work with the board.

The Tinker Board As A Desktop Computer

The most important test for a new computer.
The most important test for a new computer.

So having probed the recesses of the Asus support system for a Debian image, our Tinker Board was brought to life. After a pretty fast boot-up sequence we were rewarded with an LXDE desktop with a couple of folders containing GPIO and WiringPi software to help get started talking to hardware. Connecting an Ethernet cable and firing up Chromium gave us a respectably quick general-purpose web browsing platform that is noticeably faster than the same experience on either the Raspberry Pi 3, or the Odroid C2.

After a few minutes use though an odd problem became apparent, corruption in anything typed. This was traced by substitution of peripherals to our bench keyboard, a vintage IBM Model M. It’s the standard input device hereabouts and has been for many years, and it and a Microsoft mouse are permanently plugged into a PS/2 to USB adapter that hangs off the bench USB hub to avoid the computer having to supply the Model M’s hefty power requirement. This combo has worked with innumerable computers of all varieties over the years, yet for some reason the Tinker Board is the first to balk at it. A substitution to a USB keyboard and mouse fixed the issue, but it points to there being something amiss with the board’s handling of USB  peripherals.

The GIMP is as smooth as you'd expect it to be on your PC.
The GIMP is as smooth as you’d expect it to be on your PC.

With a working keyboard then the Tinker Board makes a very capable desktop machine. When the Raspberry Pi 3 was released last year [Brian Benchoff] hailed it as the moment at which the single board computer line could boast a credible desktop computer rather than a hobbyist’s toy, but while that remains true today a regular Pi 3 desktop user will still often find themselves reminded that they are close to the limit of their device’s capabilities.

The Tinker Board really shows its extra power on the desktop. While it will never match a high-end PC workstation you can do proper work on it without feeling held back.

How About Those Pins?

It’s very nice to have a Linux desktop, but many of us are going to be more interested in the GPIO lines. Opening up the folder containing the Python GPIO library it’s obvious that it is a derivative of the RPi.GPIO library that Raspberry Pi users will be used to. Putting together a quick test script to waggle a GPIO revealed that there was no library of that name though so after a quick run through the install process to make sure that the library was in place the name was found to be ASUS.GPIO. So starting the test script with “import ASUS.GPIO as GPIO” resulted in a script that behaved exactly as you’d expect with its Pi ancestor.

The README warns that SPI, I2C, One-wire, hardware PWM, and serial functionality is not yet implemented, but suggests that we watch that space. Given the lacklustre web support for the product it’ll be interesting to see whether they follow up on that promise.

The final point from our in-use tests of the Tinker Board comes from our temperature measurements. Our trusty radiation thermometer  failed to register anything over 27 Celsius from the SoC no matter how much effort we put it to. This compares very favorably with previous measurements from a Raspberry Pi with no heatsink at 44 degrees, and the Odroid C2 whose heatsink we measured at 37 degrees. Thus the supplied stick-on heatsink feels almost superfluous, and we certainly saw no point in fitting it to our test unit.

The benchmarks for this board and the Raspberry Pi 3 have been done to death elsewhere so there is little point in our rehashing them here. Suffice to say that it’s easy to believe a claim of the Tinker Board being twice as fast as its competitor. Detractors point at the RK3288 being a slightly older SoC than those at the cutting edge of the tablet market, but in this application its 1.8GHz clock speed and 2GB of memory is enough to leave the Pi 3’s 1.2GHz and 1GB behind. However this is not the whole story, because the Tinker Board’s Rockchip has a Cortex A17 32-bit-only processor while the Pi’s Cortex A53 is 64-bit capable. Thus it is possible that were the Pi running a 64-bit operating system rather than the 32-bit Raspbian the race between the two would be much closer.

One feature of the Tinker Board we were unable to test due to lack of an available monitor is high resolution performance. This board supports resolutions up to 4K.

Reasons To Buy One, And Reasons To Avoid

So after this extensive description of the Tinker Board experience, what might our verdict be?

For the hardware, the Tinker Board is certainly an impressive and very well-executed board. It’s quick, pretty to look at, and while it’s by no means the cheapest single board computer of its class, it’s not the most expensive either. Asus have amply demonstrated their expertise in producing high-quality hardware, and the result is a clone of Raspberry Pi hardware that’s much better than a real Raspberry Pi in almost every way. We’ll even forgive it for its refusal to deal with our thirty-year-old PC keyboard.

Editor’s Note: Please read the update starting in paragraph three of this article and take that into context when reading the following criticisms.

The community is the true added value in a Raspberry Pi.
The community is the true added value in a Raspberry Pi.

But before the Asus PR department go crazy and cut-and-paste that sentence comparing it to a Pi as a ringing endorsement, there’s a snag. There is much more than hardware to a good single board computer, we have to consider the software, and the support. This board will not only be used by Linux savants, it will end up in the hands of ordinary people with an interest in computing. Some of them could even end up in the hands of youngsters, and it is for these markets that the intangible side of the purchase is more important than the physical hardware. When you buy a Raspberry Pi, in most cases you are buying it as a vehicle for Raspbian, and when you lose your way you have also bought the Raspberry Pi forums and its community who will supply you with the answers you need to get started again.

At first glance it might seem that Asus have done a good job here, after all the Debian distribution does not have the rough edges you’d find in for example the Ubuntu Mate distribution for the Odroid C2. It doesn’t come with some of the desktop software Raspbian has pre-installed, but all of that is only an apt-get away.

Unfortunately though if you too have had to find and download the distro yourself you’ll understand exactly where Asus have failed with a capital F in cloning the Raspberry Pi. It’s understood that they couldn’t instantly replicate the huge community that surrounds the Pi, but to not even bother with a website for the product beggars belief. There’s a story about hardware companies sometimes being useless when it comes to software, as demonstrated by the lousy interfaces we see on smart TV sets. But you might expect them to look at the product they are copying and at least try to understand that there is more to it than just the board. To release such a polished piece of hardware and fail to back it up with even the most basic online support offering is a mistake we’ve seen many times before. It is one that Asus should take immediate steps to remedy if they expect this board to sell to more than a few hardware enthusiasts.

So would we recommend that you buy a Tinker Board? Of course we would, it’s a fantastic piece of hardware, and as a Hackaday reader you’re likely to be able to get the most out of it. Would we suggest that you recommend it to someone else without it being backed up by a meaningful support offering? Of course not, unless you want to become that person’s personal support hotline. Tell them to buy a Raspberry Pi 3 instead. But it may not always be this way. Asus, it’s up to you to finish this journey by providing the framework users need before they they can begin to love and adore the Tinker Board.

 

 

119 thoughts on “Review: The Asus Tinker Board (Updated)

  1. Good read, some superficial points (“fast” as a metric… how about some times?), some a bit more revealing (the catch on ASUS.GPIO).

    “but in this application its 1.8GHz clock speed and 2GB of memory is enough to leave the Pi 3’s 1.2GHz and 1MB behind.”

    Be nice, it’s 1GB, not 1MB ;)

    1. Where does one buy this board from?? It’s not on Amazon, checked ASUS’s page in both the UK and U.S. I get nothing. Even the link to the board posted on this form provides no information for purchase, I may have overlooked, please help.

    1. Open chips documentation is the key; once the manufacturer documents the chips driver will come very shortly.

      There’s a huge community of tinkerers and enthusiasts out there waiting to port software to boards like this one, but most manufacturers protect their so called “intellectual property” behind piles of NDAs, which kills any potential involvement from normal mortals.
      As an example, if ARM ltd would properly document their Mali GPU allowing open drivers to be created for mainline kernel, the Raspberry PI would be dethronized in weeks even from the only task it still performs better than the much cheaper competition, that is, accelerated video output. Sadly ARM doesn’t care about the community, so that we have really nice, powerful and cheap boards (read as more power than a RPi3 at half the cost or less) but for any use involving accelerated video we must forget about mainline kernel and rely to some years old 3.x one plus some obscure binary blob doing all the magic.

  2. The problem with all of the tiny computers is not CPU power but I/O bandwidth. Thus, I scrapped a Pi 3 for use with SDR and turned to an Odroid which supported USB 3. It appears that the Tinkerboard contains a gigabit Ethernet, it’s not clear if it’s USB 3. While the Pi has a few high-bandwidth LVDS lines on the camera and display interfaces, they rely on firmware in the Pi’s graphics processor which we don’t have proper access to or sufficient documentation to use them for anything but the intended camera and display. Just how to interface to their equivalents in the Tinkerboard is yet unknown.

    1. Does your need of I/O bandwidth go beyond SDR applications? I can’t think of many applications that need it (but that my be a symptom of its absence — nobody is building those things that need it).

      The vast majority of applications I’ve seen are using SBCs as driver for a monitor (HDMI, etc) or headless units for simple processing and connectivity. The former would be magic mirrors, video game emulators, desktop computers, and the latter is mostly camera processing (time lapse, etc.). I have occasionally seen BeagleBone Black used for lower-level hardware applications because of its PRUs. I would be interested in hearing about more I/O intensive needs.

      1. IMO once you’re trying to pull in the amounts of data that would be needed to justify requiring gigabit Ethernet or USB3 – especially for SDR – you’re probably going to exceed the capabilities of the CPU itself for most ARM boards out there.

        Exception might be some of the TI ARMs with onboard C66x DSPs – but I’m not sure if anyone has GR support for offloading to those yet. I know offloading to the GPU of a Tegra K1/X1 is very limited – gr-theano only supports 4 operation blocks.

        1. Andy, the designs I’m working on are meant to be mounted in an automobile, and fit in the form factor of current mobile transceivers. On Pi 3 we definitely run out of bandwidth. PC motherboards aren’t a good fit. The hardware design would work a lot better with a direct high-bandwidth connection to the CPU, as with Pi’s undocumented LVDS lines, rather than sharing USB with all other peripherals (and maybe the Ethernet chip too).

          1. FWIW I have an Intel ComputeStick with a core M5 processor, it has USB3… the case it comes in is about half the volume of the raspberry Pi case I have for that board. It is blazing fast for driving a 4K TV with graphics heavy games and video.

          2. Others have been pointing out various Intel solutions – I kind of agree with them here.

            There are embedded Intel solutions that are much more compact than a typical motherboard but have CPUs significantly beefier than the ARMs in the Pi or Rockchip.

            The Up! board might meet your needs – It definitely has native USB3.0 as it’s been selected by Intel for their RealSense RDK and the RS lacks USB2 fallback capability, and I’m fairly certain its Ethernet interface is also native.

            Unfortunately the Up! seems to lack CAN – which IMO is an oversight for Intel selling it as part of a robotics development kit and might impact your vehicle usage.

          3. @Nathan, intel’s computestick is 10 times the price of these boards and beyond this it lack any network connectivity so it’s totally useless! At best you can use them to block a door or something like this but it’s more expensive than a piece of cardboard for this usage.

      2. For SDR, the bandwidth requirements are unbounded. Give me more bandwidth, and I’ll digitize more spectrum. High-end HF SDRs digitize at 60 MHz, 16 bits I plus 16 bits Q. These can receive the entire HF band from somewhere in the kHz to 30 MHz at excellent S/N and, if their clocks are good, low phase noise. Some SSDs might be able to record at this bandwidth. When you get to VHF and UHF, you can use even more bandwidth.

        What else needs this sort of bandwidth? Maybe a UV SLA printer with optical galvanometers and a laser diode. I don’t know how high bandwidth the galvos can get, but it’s possible to have more than one set of galvos and laser diodes. Give it enough hardware and I/O bandwidth becomes the speed limitation.

        1. Nah, that’d be a hell of a 3D printer. And if you’ve added enough lasers and galvos, adding another CPU board is going to be cheap by comparison. If the problem of 3D SLA printing can be parallelised that much, having several lasers at once, then parallelising the computers driving it is no problem at all. Ultimately it’s just reading numbers off a list, and sending them to the diodes and galvos.

          That’s an artificial problem you’ve created there. There might be some others where IO bandwidth is important. Maybe something with digital video, or searching massive databases.

        2. “Give me more bandwidth, and I’ll digitize more spectrum”
          What about hacking those AHD video surveillance recorders? All of them already run Linux (plus proprietary blobs of course) but they’re cheap and must have really fast ADCs if they can sample 1080p video at 30 fps or more.

  3. rk3288 kick the ass of other SOCs. for pure performange is te top, but im afraid the board nne community to reach bettert lever about software. anyway im sick of all mainline hardware as arnuino and raspberry, i prefere semi-unknown board like this.

  4. Looks like a pretty credible alternative to the Raspberry Pi, the Breadcom SoC on the Pi has many limitations that this seems to address. Such as the 1GB memory limit and the horrible ethernet via USB. The 4k support & digital audio outputs should also make this popular with people building media players.

    The price seems pretty good as well considering all the extras you get for the extra £20.

  5. Linux version 4.4.16-00097-gc9727d4-dirty (yihsin@yihsin-System-Product-Name) (gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) ) #23 SMP Fri Jan 13 12:29:56 CST 2017

    Where is the source code?
    Where is the rest of the schematics?

    1. If you are referring to my post:
      I was there. That’s why I know that there is no source code and that the schematics are incomplete.
      Or did you really believe the RK3288 has only two ground pins?

    1. Please use the “Reply” link to post a reply.

      I’m talking about this single page schematic. Only the P26 and P27 ground pins of the RK3288 are shown.
      Are you aware that the RK3288 has 636 pins? Do you see the RTL8723BS Wifi module, the ALC4040 audio codec, the RTL8211E ethernet phy, the GL852G USB hub or the two K4E8E324EB-EGCF LPDDR3 memory chips anywhere in the schematic?
      This schematic is a joke.

      The Firefly Team hosts lots of interesting files about the RK3288. I think I’ll create a mirror on my hard drive just in case…

  6. Side note on that “trusty radiation thermometer”: Infrared thermometers aren’t reliable when pointed at shiny metal targets! If the emissivity of the surface isn’t reasonably close to 1, it’s acting more like a mirror, and you’re taking the temperature of the rest of the room reflected off the surface of the shiny thing.

    Emissivity of many materials (wood, leather, brick, most plastics for example) is in the 0.85-0.97 range, and they work well with infrared thermometers. Emissivity of most bare metals is usually 0.3 or below, fuhgeddabouttit!

    The simple remedy used by thermographers is to stick a piece of tape (matte black masking or gaffer’s tape is ideal) onto the shiny object and point the sensor at that. Mind the distance-to-spot ratio and you should be golden.

      1. And everyone’s favorite hackers from the Netherlands already established: If it smells like chicken, you’re holding it wrong!

        To the points farther down about throttling, I wonder what the thermal resistance of the package is. If the outside were kept quite a bit cooler than room temp, would the innards of the chip be free to run faster? i3Detroit has this LN2 dewar, and if we can get our hands on a board…

  7. I don’t think your temperature measurements are correct. My single board computers, including an O-Droid, Arndale, IFC6410s, Raspberry Pi 3, run MUCH hotter than that. For apples-to-apples, I installed a medium sized heat sink on my Raspberry Pi 3 and it still hit 80C under moderate load (with the governor set to Performance). I had to change the governor to Power Save to keep the heat down.

    You can’t measure the metal caps of those chips accurately with a laser thermometer. The emissivity of the metal cap prevents anything close to an accurate reading. Also laser thermometers have a pretty substantial measuring “cone” for lack of a better term. It’s like 20 degrees or something. You can’t stand 3 ft. away and put the red dot on the chip because it will average a large area and also return a very misleading result. Put a piece of tape over any shiny metal surfaces if you want to take a temperature measurement.

    Also your chips are almost certainly throttling to achieve those low temperatures under load. That pretty much makes all thermal and performance measurements meaningless if the chip is limiting itself internally to stay within a thermal envelope.

    1. In fact, all of these boards (including the Tinker and MiQi) throttle by default above a certain temperature. Since the RK3288 is very fast, when it throttles at 1.2 GHz or so, you don’t necessarily notice immediately and it quickly recovers. But for my build farms I had to install some large heat sinks to save them from throttling. Unfortunately the Tinker board doesn’t leave enough room to place a large heatsink so it will always throttle under load.

      1. THIS!

        PEE 3 throttles so much out of the box its not even funny, what is funny tho is throttling being hidden in the gpu binary blob and not informing linux kernel about the actual real MHz data -> you dont even notice throttling unless you run benchmarks/stress tests.

        Apart from throttling there is also Chinese culture of Lie. We had a rather big stink about Amlogic (yes yes, you will say company is in US, but it sells exclusively to chinese market) S905 claiming to run at 2GHz when in reality any setting above 1.5GHz was being ignored. It took over a year for someone to finally run proper tests and notice.

        http://www.cnx-software.com/2016/08/28/amlogic-s905-and-s912-processors-appear-to-be-limited-to-1-5-ghz-not-2-ghz-as-advertised/

        37 degrees is suspiciously cool for a 2GHz 4 core chip.

        1. I noticed it out of the box as soon as I started a build test, actually I was the one reporting it there :-) At first I thought the CPU was only throttling and it took me a while in kernel code inspection to see that I could find nothing and that it had to be deeply buried in the device. Due to this, I don’t trust *any* amlogic-based device anymore. While with allwinner you never know what you’ll get because apparently the chip vendor relies on customers to find the highest frequencies by themselves in field, amlogic lies to customers and that’s even worse. S912 would be nice at 2 GHz, but who will believe their claims now ? Rockchip’s kernel “only” apply their “SAFETY_FREQ” limitation in their own kernel that’s trivial to unlock and that doesn’t exist in mainline, so you can have all the frequencies you want using the device tree.

    2. OK, forget the radiation pyrometer, if you stick your finger on it it’s just warm to the touch, not hot like the Raspberry Pi 3. As you might expect, this isn’t the first processor temp measurement I’ve made :)

      1. It’s probably worth outlining the kind of things I did to get the measurements on this thing. I’m trying to get a general-purpose user experience, because I suspect most people owning a Tinker Board will not for instance compile a kernel. So I’ll try to get as many big resource hungry processes my real-world user might run, only simultaneously. Playing a video, unzipping a large file, things like that. The same load lit up the Raspberry Pi 3 like a Christmas tree, but left the Tinker Board merely warm to the touch. What else can I say.

        1. I agree to a point. The problem there is the chips realistically are all cell phone chips, and they throttle. So really you’re only measuring the temperature setpoint at which they are programmed to begin reducing their performance. You can probably get that from a datasheet or source code – no measurement necessary. So performing these tests in a non-cell phone environment is of limited value. That might give you an indication of performance per watt (assuming each board will naturally radiate ~1 watt), but total performance is obscured. So yes, a 13 year old who is a beginner, plugs the board in and inserts an SD card to make a light blink, these measurements will apply to them. But anyone doing anything more advanced is going to hit a wall. I would venture to say that most people on this website are probably asking a lot more of their boards. Particularly if they have one hooked up to a keyboard and monitor and are using it as a desktop replacement. Nobody wants to watch a Youtube video and have their computer throttle.

          To set a frequency governor in Linux you don’t need to compile the kernel. It’s actually super easy to do. If you’re running a graphical environment you can usually just click a button to change the frequency scaling, but if you’re running console (all of my SBCs are headless) then you just have to run one of these commands:

          cpupower frequency-set -g performance
          OR
          echo “performance” > /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

          That’s assuming you want the performance governor. There are other governors to choose from, depending on your application. I had to limit my Raspberry Pi 3 by setting it to “powersave” because it kept overheating. That’s with a decent heat sink, sitting in a well ventilated enclosure, outside with good airflow and not in direct sunlight. And really each core is probably only sitting at 50% utilization if you average it out. Even with all that the thing still got wicked hot.

      2. The sensation of temperature is entirely dependant on the temperature of the part of your body you use at the time of probing. Warm to the touch in california could be blazing hot in scandinavia. Fingers are the worst possible part of your body for these kind of measurements as their temperature varies alot. You should either calibrate your hand to a known temperature everytime you probe or use a more temperature stable bodypart.

  8. For those who can’t wait or who want some on-board storage for the same price, mqmaker’s MiQi board has been around for about 6 months now, is very stable and uses the exact same components except that in addition is features some eMMC storage on board that the Tinker board doesn’t have. It works fine with the mainline kernel (the Tinker board will also work with it) so there’s no reason to keep the default bogus vendor kernel. I’m using them in a build farm, and these board are impressively fast! The RK3288 at 1.8 GHz is on average twice as fast as the RPi3. Oh by the way, with the default kernel, these boards are often locked to 1.608 GHz while claiming to run at 1.8, but it’s easy to notice in benchmarks. With a mainline kernel it’s trivial to unlock them (my MiQis run at 2 GHz). More info here :

    https://forum.mqmaker.com/t/miqi-based-build-farm-finally-up-and-running/605/24

    Apparently the really nice thing Asus has managed to do is to have the *exact* same connector positions as the RPi so that it should be able to reuse the same enclosures. Most other boards often vary by 1-2 mm and cannot smoothly reuse them. It might be that eventually the RPi form factor will become a de-facto standard like the ATX and ITX form factors were standard for PC board makers.

  9. For those who can’t wait or who want some on-board storage for the same price, mqmaker’s MiQi board has been around for about 6 months now, is very stable and uses the exact same components except that in addition is features some eMMC storage on board that the Tinker board doesn’t have. It works fine with the mainline kernel (the Tinker board will also work with it) so there’s no reason to keep the default bogus vendor kernel. I’m using them in a build farm, and these board are impressively fast! The RK3288 at 1.8 GHz is on average twice as fast as the RPi3. Oh by the way, with the default kernel, these boards are often locked to 1.608 GHz while claiming to run at 1.8, but it’s easy to notice in benchmarks. With a mainline kernel it’s trivial to unlock them (my MiQis run at 2 GHz). More info here :

    https://forum.mqmaker.com/t/miqi-based-build-farm-finally-up-and-running/605/24

    Apparently the really nice thing Asus has managed to do is to have the *exact* same connector positions as the RPi so that it should be able to reuse the same enclosures. Most other boards often vary by 1-2 mm and cannot smoothly reuse them. It might be that eventually the RPi form factor will become a de-facto standard like the ATX and ITX form factors were standard for PC board makers.

  10. To be honest, the RPi forums do not do a good job at helping me.

    The last thing I posted had the full source for a minimal program that did not work correctly using a library on the Pi, no responses at all. Been half a month.

  11. I have one of the “Leaked” boards from CPC, according to twitter the board was pull from their site because of software issues. However, it can be bought from another high street retailer http://www.currys.co.uk/gbuk/computing-accessories/components-upgrades/raspberry-pi/asus-tinker-board-10158412-pdt.html

    Their PR department is pretty useless tbh

    With high load it was quite easy to hit 75c+ on this while compiling retropie and the included heatsink and a small fan is a must really if you want to load the cpu.

    And just as this post has no videos here are few tests I’ve made with the first image released.



    1. 75C sounds realistic, almost makes me think Jenny either didnt test this board at all, or is a cyborg with plastic fingers and writes reviews as a way of infiltrating meatbag subcultures.

      1. My MiQi boards remain totally cool with a single core at full speed, and it’s not that easy to use 4 cores with regular workloads so I think he tested it correctly and that there was no issue with this.
        BTW, 75°C is the critical point where throttling starts in the rockchip kernel.

        Those interested in making this card heat for real can test this one, it’s impressively efficient by running some carefully optimized crypto asm code :

        while : ; do openssl speed rsa2048 -multi 4; done

        That’s how I test my boards when overclocking. Nothing comes close in terms of power consumption and heat dissipation. I reached 4 amps over the micro-usb cable at 2.1 GHz vs 2.5 during kernel builds! Believe me, it gets hot even with a heatsink! Oh and at 1.8 GHz it should report ~405 sign/s, vs 360 at 1.608 GHz, so it’s trivial to see if the kernel cheats :-)

        1. I was logging the frequency and I did notice is start to throttle down to 1400 and eventually power off due to a cheap “2.1A” phone charger.

          4 amps over a micro usb is impressive! The ASUS documents advise not to overclock and I did noticed that in their init.d script it was limiting cpu to 1400~ rather that its quoted 1800.

          Either way with rockchip still supporting the 3288 and with their rootfs I think this might actually be a pi 3 killer soon..

          1. You definitely need a very good power supply to reach 1.8 GHz. The reason for 4 amps in my tests were the micro-usb connector. Above about 2.5A, no less than 0.5V are lost in it, so the DC/DC regulator has to pull even more juice to compensate, resulting in something like 4.5V/4A at the DC/DC input for 5.2V/4A out of the power adapter, the difference being lost in cables and micro-usb connector. That was not fun.

            I’ve got many outages due to bad power supplies but once you have a really good one (aim at twice what you need to be satisfied), the boards are very stable. The CPU doesn’t like running above 92-95 degrees however. The DTS throttles above 70 (mainline) or 75 (BSP) but you can push it a bit forward to avoid needlessly throttling. But without a heatsink it’s just a matter of time :-)

        1. ANY load lights up the PI like a Christmas Tree.

          Have you forgotten Gareth Halfacree’s post to the world re the Pi3’s meltdown mode, running at close to 100C for no good reason? Yes? That means the Raspberry Pi Foundation’s tactic of sweeping bad RPi information under the rug is a phenomenal success.

          Don’t feel too bad; most people, in a spate of tribal morality and insanely-misplaced loyalty forget ALL the negatives associated with the RPi.

  12. This thing was reported by others (like all of them) back when CES was still going on. Hack-a-Day is so far behind the times. And yes, it’s not officially out yet. In fact, you can’t find hardly anything about it as ASUS has pulled everything down until they give it a proper release.

    I was excited about it until seeing it’s only 32bit. Why didn’t they go with a faster chip AND better bit width? I tire of tradeoffs. Just gimme the thing loaded and let me decide what I want to pay for it for once.

    1. And why precisely do you need your CPU to be 64-bit ? The current 64-bit chips are either dog slow (cortex A53 like in Rpi3, Pine64, orangepi-pc2, odroid-c2, etc) or very expensive and power-hungry A57/A72. Also most thumb2 code (armv7) runs faster than the equivalent armv8 code on these CPUs. The Cortex A17 gives you the best of both worlds, a very fast CPU core, a fast *dual-channel* memory controller, a reasonable power draw and a reasonable price. At the very moment thanks to its A17 core, the RK3288’s performance/price ratio is unbeatable. Atoms are a bit more expensive and and a bit slower, all other 32-bit ARMs are slower, and 64-bit ones are either slower or more expensive.

    2. It doesn’t really make a difference. 10% maybe? If it had more than 4 gigs of RAM there might be a case, but currently these boards don’t really pack that much memory. Everything fits nicely into a 32 bit address space.

  13. yeah 2 years late there is a LattePanda board windows 10 on a quad core 4gigs ram 64 gig emmc …. i have helped kickstarted this project, running very well no problems after 6 month of use

  14. I had a lot of trouble with ASUS support quoting me one price for a repair, and then multiplying that by 3 when I shipped it to them, and refusing to sell me repair parts to do it myself. I can only imagine their support for a DIY hardware platform like this will be similarly dreadful.

  15. Now when first learning of the goodness of Pi of the raspberry variety, and observing a little difficulty in a reasonable way of getting hold of one… (Qty 1 with $20 shipping = nope) … a little whisper started in my head, orange pi, orange pi… and as it seems increasingly likely that rPi will never ever ever be available in quantity, and shipping it got louder…

    Now seeing stuff like this it’s yelling ORANGE PI, ORANGE PI and even getting as bad as this…

    1. Apologetically make for the half sensy wordy stringing. Sleep of three hours only having yesternight. I meaned the pi zero in particular, and 1th para, last sentence should word eye as “and shipping costs not any better, it got louder… ”

      Meaning, some suppliers seems like you can order an arduino, mini, or clone, 5 of them even, and have ~$7 shipping but it’s double for a pi zero. They need the xenon filled bubble mailers obviously.

  16. Hi, one thing I really like: the RK3288 has a Mali T760 GPU, which supports opencl, the framework to run generic (parallelised) applications on the GPU.
    I may be wrong (feel free to correct me if I am wrong) but I think this is the first relative cheap linux board I see that has a GPU that supports opencl. Hooray! :-)

      1. Yes, it has. Kernel side is open source since years and user space library is even on Arm site under Firefly board (uses same SoC). If you search a bit, you can even find more variants, I know at least wayland and fbdev.

  17. I managed to acquire one about a week ago. Couldn’t get it to boot with any of the USB power adapters in the house. The closest was with an Apple iPad charger. Purchased an official Raspberry Pi 2.5A supply and it now boots without a problem. Have had problems with two wireless keyboards (Microsoft and HP). Keys either don’t take or repeat like crazy.

    Main reason for the purchase was the claimed high-def audio (I want to record bats). I need to dig out another keyboard and mouse and will have another play tomorrow.

  18. You mention other comparable boards apart from the Pi 3. In your opinion what is the best similar priced boards for downright CPU speed ? If I want to number crunch on a robot, whats my best bet ? I’m already using 6 Pi3s, 2 Pi zeros, a banana pro, Intel DX2000, 4 X Intel Arduino 101, 8 Arduino Uno and Nano etc. So what I want now is something that can crunch AI data.

    1. For me, it’s by far mqmaker’s MiQi. It’s the same CPU, it runs at 1.8 and can even be overclocked to 2.0 if you have a good PSU and correct cooling (large heatsink but 45×45 fits on the board). I use it in my build farms, 4 MiQi are twice as fast as my core i5-3320M at 3.1 GHz, and only costs about $60 or so, and includes the eMMC storage (hence no moving parts)!

      Then if your workload is heavily threaded, you can use FriendlyARM’s nanopi-M3. It contains an octo-A53 running at 1.4 GHz without throttling with a reasonably small heatsink, and is only $35. With its 8 cores, it’s almost as fast as the MiQi with its 4. It’s the one I use on a walking hexapod with computer vision. However keep in mind that while belonging to the ARMv8 family, the kernel is only 32-bit there so you will have to run it in 32-bit mode (often not a problem unless you needed some armv8-specific optimizations).

      The odroid C2 is really not bad at all and keeps cool even at full load. The Up-board featuring a quad-core Atom x5-z8350 at 1.92 GHz (up to two cores) or 1.68 with 3 or 4 cores running is pretty fast. Not as fast as the MiQi, about 10-20% slower only when using all 4 cores, and approximately the same with only 1 or 2 cores. However you have SSE and various other x86 stuff that may allow you to use certain libraries that were already optimized for this architecture. It’s a bit more expensive but not that much (about $90).

      Hoping this helps.

    2. what do you mean crunch AI data? do you mean CNNs for vision? crunching is the part you do on a desktop with couple GPUs, the resulting trained neural network is quite light on computation.

      anyway the best performance for buck and watt you can get is a 2-3 year old Laptop with a broken screen off ebay/friend. You should be able to find ~$100 Sandy Bridge quad core (8x 2.8GHz 64bit) machine with cracked lcd and no hdd if you look hard enough.

  19. @Steven Clark:
    “RTC and power management sound exactly like the kind of things the Pi is missing.”

    ONLY RTC and power management? Good choices for starters, but how ’bout

    1.8 GHz clock, upgradeable to 2.0 GHz without breaking a sweat–or catching fire; (The final point from our in-use tests of the Tinker Board comes from our temperature measurements. Our trusty radiation thermometer failed to register anything over 27 Celsius from the SoC no matter how much effort we put it to);
    REAL GHz ethernet;
    4K video;
    2 (TWO) GB of memory;
    a machine that’s not absolutely CRIPPLED by running everything through a a USB chip;
    total ignoring of any RPi shortcomings…

    Jump right in and add your own, folks; there’s plenty of opportunity.

  20. I have one of the leaked boards and so far it does appear quite quick with regular desktop activities. I have a few PI’s and have some dedicated to different tasks such as NAS, Backup or Kodi.

    The Tinker Board OS at the moment is the barest of bare. Hardly anything works at all and the rest needs significant messing around. The only thing I have working straight off the bat is Webmin and just to test it. BOINC works on command line only. LibreOffice installs and appears on the desk top but does not run. OpenMediaVault has too many failures to install properly. I have tried others.

    Even the Kodi image disappeared from the ASUS website.

    There is a lot missing from the OS at the moment. Yeah, I got one of the cheeky leaked boards, but at the moment it can not do a fraction of what a Pi does.

    1. Look at the mqmaker forums. The board is pretty much the same (except that it’s been shipping for more than 6 months now), it works fine with mainline kernels and there are people like Myy maintaining their own patchsets to improve graphics support or things like this (sorry for not being accurate, I do not use the graphics stuff at all).

    1. what browser? sunspider is more of a javascript engine test than a platform
      anyway my $50 laptop board from ebay using Vivaldi(=chrome) scored
      stock 2.6GHz Total: 281.9ms
      OC 3.1GHz Total: 231.5ms

      Thinkpad x230 motherboard with i5-3320m from a borked laptop = $50 (x220 i5-2520M ones were $20). Sure, its >2x the size and has a weird shape, but cant argue with bang for the buck. I also bought $15 one with non working internal keyboard circuity (smc or connector), usb keyboard works fine, this one will end up as a NAS, I hope I can get SATA port multiplier working on it.

  21. I bought an ASUS Tinker board from Amazon USA on 2/23 and received it a couple of days later, I was able to download a Debian image from the ASUS UK site and create a bootable microSD card that did boot properly.

    That’s all I have had time to do with it.

  22. After a bit of hunting following the ghost launch I managed to locate one in Germany, but it seems the supply is starting to ramp up and they are now flowing into the market.

    The support site is still generation 1 and nothing by way of a community, but I was able to locate and download Debian. Sadly at work, using a Pi power unit it refused to boot, however once home with the same image and a beefier PSU it fired up in no time. It is nippy and responsive on the desktop, and after enabling its auto network, it went on line and happily performed and update/upgrade

    It is a no frills standard Debian, it resized my sd card itself, and logged in to a sparse gui with nothing notable to report, except for one amazing features. It has OpenGLES3.1 libs on board. I’ve not had a chance to try them out yet, as I’m a bit busy but if this is indeed a properly implemented OpenGLES3.1 system, its hands down going to beat almost all the other SBC’s out there.

    CPU gets mighty hot, so don’t try to run it without the heat sink, and as I say it needs a good power unit.

    So far so good. I can’t wait to get programming some OpenGLES3.1 on it and see what it can really do.

    1. As a small follow up, I’ve had a chance to play with it now, and CPU wise its indeed a beast and the latest tinker os is stabe, but I’m very unhappy with the GPU performence, GLMark2 produces a woeful rating of 44, lower even than my Mali400Mp2 powerd NanoPi M1. It would seem that despite the header files being on board all current GPU functions are emulated (VBO and non VBO functions taking the same time being a major clue).
      Hopefully a prompt update of he IS with hardware drivers will be a priority for Asus, since there’s no point in producing a board with this powerful a GPU and not acually opening it up for use.

      1. dont count on any drivers, you have been warned in comments above
        there ius a reason so many people go for PEE, its the only reasonably supported arm soc dev board out there (other than i.MX ones)

        1. Completly understand that, but its Asus, not some small warehouse shipper in Korea, this is one of the biggest makers in the world, incluidng a lot of graphic cards, so if they have come to this market they must surely have a means to get proper drivers set up, otherwise they bcould have sourced any cheap rubbish Mali400 based unit and chucked them out.
          I have no inside info, but I just can’t see a company that size not supporting a board this powerful and going after the market that Pi has.
          .

          1. Well I’ve been around the block a few times and seen Asus screw users over. Never know what their attention is going to be focussed on. For example in the PIII era, you would have gotten kickass excellent support for their 440BX based boards, for their Via based ones, yah, generic flaky via 4 in one driver and the best of luck to you.

          2. Well I guess time will tell, I like to collect SBC’s so if nothing else it adds to my collection, but if drivers come along it could be a main target. Even if emulated the chance to play with OpenGLES3.1 on such a device is fun.
            We all buy these things for a range of reasons and they all fill difference niches in the market, so I will take this on its own merits and hope it lives up to its potential.

  23. The Tinker Board is back on Amazon for $59.99, with free shipping!
    https://www.amazon.com/ASUS-Tinker-board-2GB-Motherboard/dp/B06VSBVQWS
    Asus’ US web site does not have a Tinker Board page yet, but the UK & Philippines do:
    https://www.asus.com/uk/Single-board-Computer/TINKER-BOARD/
    https://www.asus.com/ph/Motherboards/TINKER-BOARD/
    These accessory boards will fit:
    http://www.tinyosshop.com/index.php?route=product/product&product_id=971
    https://www.amazon.com/Prototyping-Shield-Raspberry-Model-Version/dp/B01K9XR6BI
    http://www.tinyosshop.com/index.php?route=product/product&product_id=774
    https://www.amazon.com/Prototyping-Shield-Raspberry-Pi-Model/dp/B00N769618
    This is the graphics chip on the board:
    http://www.realtek.com.tw/press/newsViewOne.aspx?NewsID=342
    “The Realtek RTL8723BS is an SDIO 2.0/3.0 solution comprising WLAN, BT, and FM Receiver. Its small module size, low power consumption, and intelligent coexistence mechanism makes it ideal for multiple applications such as tablets equipped with the latest Intel Bay Trail platform, smart phones, 2-in-1 detachable/convertible notebooks, and HDMI dongles (TV stick).”
    So it has a build-in FM receiver! Does anyone know if Asus has made it active? If not, let’s encourage them!

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