Asus Tinker Board gives the Raspberry Pi 3 a run for its money

Most would say the original Raspberry Pi got the single-board computer craze (SBC) going. Since then, the Raspberry Pi got three major hardware revisions, and the market at large has come up with similar devices in countless flavors, including Orange and Banana. According to Hexus, Asus decided to throw its prodigious engineering muscle into the SBC ring with the cheerfully-named Tinker Board, which one-ups the Raspberry Pi 3 with significant connectivity advantages, double the memory, and higher clock speeds.

Image source: CPC

The Tinker board is powered by a Rockchip RK3288 SoC based on an ARM Cortex-A17 design and running at 1.8 GHz. The Tinker's heart lacks the 64-bit support found in the Pi 3's Broadcom BCM2837 Cortex-A53 chip, but the performance figures Hexus got a hold of show a 3,925-to-2,092 score advantage in GeekBench. Some of that performance difference might be from the 2GB of dual-channel LPDDR3 memory in the Tinker board, versus the Pi's 1GB of DDR2 at 450Mhz. Asus' board includes 802.11n Wi-Fi and Bluetooth 4.0 support. Storage options appear to be limited to microSD cards, just like the Pi.

Those interested in an SBC for a home-brewed media playback machine will like the Tinker's 4K playback support over the HDMI port, via its SoC's Mali-T764 GPU. Asus says its audio solution is superior to the Pi's, with support for 24-bit samples at rates up to 192 kHz. We'll have to wait for third-party testing to hear if the Tinker's analog audio output is an improvement over the Pi's poor singing.

The Tinker Board has the same 3.4" x 2.2" (8.5 cm x 5.6 cm) dimensions as the full-size Raspberry Pis. The 40-pin GPIO header is in the same physical location, but the pinout is unique. We were unable to find official power consumption specifications, but we expect that buyers will need to provide more juice than they would for a Pi 3. In any case, software support is generally more important than hardware in ARM-based SBCs. Asus says the Tinker will support Debian Linux and that the device will be able to run the Kodi media software.

The Tinker Board is available now from several European resellers, though prices are not consistent. We reached out to Asus for a US release date, and we're told to expect the Tinker Board to hit US shores on January 30. The device appears to be selling for £46 before VAT in England, or about $57. As a comparison, the Pi 3 goes for £33 plus VAT in England and $35 in the US.

Comments closed
    • evilpaul
    • 3 years ago

    Does the CPU actually run at 1.8Ghz or is there a binary blob for the kernel that forces it down to 1.5Ghz like most of the AMLogic SoCs that are supposed to run at 2.0Ghz and I think some of the other Rockchips that don’t run at their specified clocks either.

    • NTMBK
    • 3 years ago

    This is the same SoC that is in the GPD XD, an Android portable gaming device. Quite a few videos on Youtube of it running various emulators, if anyone is interested in performance.

    • notfred
    • 3 years ago

    There’s also the PINE64 boards [url<]https://www.pine64.org/?page_id=1194[/url<]

      • willmore
      • 3 years ago

      Which seem nice, but have their own set of problems.

      If someone looks for a good board that’s better than the Rpi 3, has a good community, commercial support, and good specs, check out the ODROID C2.

      The Orange Pi boards are nice and cheap and have decent specs, but they have near zero support from the vendor. There is Armbian for most of them and that’s a nice community.

      In particular, the Orange Pi PC2 board looks promising, but the Armbian port isn’t ready yet. Once it is, it’ll be based on mainline linux (4.10) and support the full 64 bit environment the chip is capable of. Also, it’s $15 plus shipping. So, roughtly half the price of the Rpi3.

      If you want cheaper and small, there’s the Orange Pi One or the Orange Pi Zero–that one is small, has wireless built in, and costs $9. There is a 256MB version for $7 if you have a project that fits in that little memory.

    • davolfman
    • 3 years ago

    I don’t give a damn unless the driver support is absolutely fantastic. Unless it can be booted with U Boot and the mainline kernel (and a few very portable blobs) it’s going to have serious trouble staying as useful as a Pi.

      • willmore
      • 3 years ago

      Mainline on the Rpi would be nice. Heck, 64 bit would be great. More than 1GB of memory wouldn’t hurt. Native ethernet would be nice, too.

      The community can’t fix hardware issues.

    • DragonDaddyBear
    • 3 years ago

    This thing is not only more powerful but it has a gigabit NIC. I wish there was a way to hook up some storage that is faster than the USB 2 to make it a NAS.

      • backwoods357
      • 3 years ago

      Banana Pi M2 Ultra?
      Pain in the ass to find a store carrying it, except amazon.
      [url<]https://www.amazon.com/Banana-Pi-Computer-Octa-Core-Ethernet/dp/B018IN6FZE/ref=sr_1_1?s=pc&ie=UTF8&qid=1485115433&sr=1-1&keywords=banana+pi+m2+ultra[/url<]

        • Anton Kochubey
        • 3 years ago

        Eight A7 cores is just stupid. An A72 dual-core would be better performer for 90% of the tasks. Other than that, that’s a nice SBC, yep.

          • tipoo
          • 3 years ago

          A dual core A72 would also fit a lot more than 8 A7 cores in. A7/A53 are so remarkably small people just sprinkle them in like salt.

            • Musafir_86
            • 3 years ago

            Welp, friends, it uses Cortex-A17, not Cortex-A7. Cortex-A17 is better than Cortex-A9, more efficient than Cortex-A15, and it’s the lastest, best perf/watt ARMv7-A core ARM ever designed…

            Regards.

            • tipoo
            • 3 years ago

            Plus I think the A53 is the highest perf/watt 64 bit processor out there afaik, so ARM has the win on both. Impressive stuff.

            • willmore
            • 3 years ago

            I think that title is held by the A35, but it’s not made on older processes.

          • willmore
          • 3 years ago

          The problem here is the cheap fab processes these SoCs are made with. You’re not going to get an A72 core on 40nm. ARM simply doesn’t support it. A7? yes.

          Now that you’re on 28 or better fab process, guess what? Your chip costs a lot more. Now you’re not able to compete in cost with the A7 chip. You’re in a different market segment.

            • NTMBK
            • 3 years ago

            Is there a single 40nm eight-core A7 SoC? I thought that they were all 28nm?

            • willmore
            • 3 years ago

            Yeah, the Allwinner R40/H8/A83T (likely one chip with different bond outs) is an 8 core A7 and is believed to be on 40nm.

            It’s used on the Banana Pi M2 Ultra that backwoods357 mentioned earlier in this thread.

            It was a complete failure in the market.

            • NTMBK
            • 3 years ago

            According to the Allwinner press release, it’s 28nm. [url<]https://web.archive.org/web/20160316222425/http://www.allwinnertech.com/plus/view.php?aid=457[/url<]

            • willmore
            • 3 years ago

            Oww, good catch.

    • NTMBK
    • 3 years ago

    These boards are getting pretty powerful. Quad core, 3-wide out of order at 1.8GHz… That’s a lot of performance for such a tiny board.

      • willmore
      • 3 years ago

      Yeah, that’s up in ODROID XU4 range. Beefy.

    • Ninjitsu
    • 3 years ago

    I question the value of running geekbench on something like this.

    inb4 “Can it run Crysis?”

      • Srsly_Bro
      • 3 years ago

      you just brought it here.

      • ptsant
      • 3 years ago

      Actually, with dosemu it can probably run a ton of legacy games, providing fun for the whole family.

    • Grumpy Old Man
    • 3 years ago

    Having a “Unique GPIO”, it sounds like pre-existing hats might not work, and requiring more power may make finding a standardized power supply may be an issue. A faster chip and more ram is nice, but I’d rather have 64 bit chip, and a 64bit OS that doesn’t time out after a 60 day eval period.

    Long story short, it sounds like a nice Kodi board.

      • Eversor
      • 3 years ago

      You can buy the 2.5 amp official Pi PSU. If numbers are true and it is 5W tops, no issue unless you try to plug USB powered HDDs.

        • willmore
        • 3 years ago

        The Rpi3 powered over the micro-USB connector can’t run full load without clocking down.

        Stop using micro-USB for power. I know it’s convenient, but it’s a bad idea.

          • Chuckaluphagus
          • 3 years ago

          You made this same claim in a previous thread about the Raspberry Pi 3, and I provided evidence from my own tests ([url=https://techreport.com/news/30975/updated-pidrives-give-tiny-computers-more-affordable-storage?post=1009126#1009126<]running sysbench with 4 threads[/url<]) to the contrary. You didn't provide any sort of satisfactory rebuttal, just stated that I should download a different stress-test program from github. It also doesn't make much sense to me that the Pi would overheat and throttle down from one power input but not another - why would the heat output of the [u<]CPU[/u<] be dependent on how you're supplying power? Insufficient power should lead to instability, not overheating. If you can present some data to back up your claim, that'd be good, but you've yet to substantiate your blanket claim that the Pi 3 overheats using the default microUSB power input, especially in light of contradictory information.

            • NTMBK
            • 3 years ago

            Perhaps some part of the power delivery circuitry starts to overheat, and triggers throttling?

            • willmore
            • 3 years ago

            Nope, it’s the brownout detector seeing too low of an input voltage. This triggers an interupt to the VPU which throttles the ARM cores.

            • willmore
            • 3 years ago

            I explained the issue sufficiently for you to reproduct it. If you’re too lazy to download a program from github, then you don’t seem to have much interest in understanding the issue.

            The problem comes from the contact resistance of the micro-USB connector. At the current the processor uses for full load the input voltage due to the resistance of the connector before what the brownout detector trips at. This causes the processor to get clocked back.

            Again, it’s not a thermal issue. This happens instantly when the current increases and well before the chip has any chance to heat up.

            I’m not going to bother refuting your strawman. It has nothing to do with thermal issues.

    • stdRaichu
    • 3 years ago

    Well this one is right up my street so I’m thinking about ordering one (before my GBP becomes completely worthless anyway) in the hope that it’ll do what my RPi’s couldn’t.

    I’ve been (trying) to use a small webcam and plain-jane [url=https://en.wikipedia.org/wiki/Motion_%28surveillance_software%29<]motion[/url<] as a pseudo-doorbell (with an eye to hooking up a screen that turn on automagically whenever anyone approaches the door, and eventually relaying a mobile call or text through a GSM modem). Doesn't need a whole pile of CPU but does need some half-decent IO and... well, sad to say but IO on the Pi is pretty crap; half the time the ethernet would crap out sending the movie stream to the server. There's a connector on the Tinker board that looks the exact same as the RPi camera module (which is thoroughly excellent, especially the newer one in low-light) so I'm [s<]quietly[/s<] quite loudly hopeful for this one. The rockchip stuff has generally been fairly OSS-friendly TTBOMK. PoE would have been the icing on the cake, but SBCs with inbuilt PoE are typically at least an order of magnitude more expensive (and yes I have been looking... especially at you Gateworks!), so may as well just use a PoE splitter. Will wait for a block diagram and the first rumblings from the community before I pull the trigger (oddly enough no word on the Asus website yet, and their support for open source in the past has been lackadaisical at best) but am wishing this will become one of the better supported SBCs out there... as at least DrDominodog51 and srg86 have observed, these things survive by dint of the people who use (and fix) them. Fingers crossed for a bit more healthy competition in the ARM world.

    • PBCrunch
    • 3 years ago

    I really like the color-coded GPIO pins. That could make things a lot easier if you have to connect GPIO stuff in a poorly lit environment.

    • ozzuneoj
    • 3 years ago

    For some reason, when I saw Asus I was really hoping this would be x86.

      • Buub
      • 3 years ago

      ARM is generally a much better choice for these applications.

      • tipoo
      • 3 years ago

      Apart from running full fat 64 bit Windows on this (why?), why are you hoping for that? Now that Windows on ARM gets 32 bit x86 emulation, and ARM chips in the low end have been beating Intel on performance and efficiency for a while. Intel doesn’t even make the line of Atoms that this would have gotten anymore. And if it was Core M, the price would shoot up to hundreds.

      • Ninjitsu
      • 3 years ago

      You can get x86 “maker” boards straight from Intel (Galileo, Edison, etc.).

        • willmore
        • 3 years ago

        And they suck and nobody wants them.

      • Luminair
      • 3 years ago

      I’m going to blow your mind right now: [url<]http://www.aaeon.com/en/p/up-board-computer-board-for-professional-makers[/url<]

    • DrDominodog51
    • 3 years ago
    • DrDominodog51
    • 3 years ago

    A SBC is only as good as its community.

      • just brew it!
      • 3 years ago

      Asus is clearly aware of that; otherwise they wouldn’t have taken pains to clone the physical layout of the Raspberry Pi 3. The intent is to be a more powerful drop-in replacement for the RPi 3, leveraging the existing community.

        • ptsant
        • 3 years ago

        That’s nice, but the enclosure is a $10 affair. Software is the real difficulty. Can you boot the same system on this one? Do you need to recompile everything and the kitchen sink? Anyway, competition is good.

    • tay
    • 3 years ago

    My wallet is ready.

    • derFunkenstein
    • 3 years ago

    Not only is this the same size physically as the RasPi 3, but the port layout looks identical, too. If that’s true, it should let this mini PC drop right into existing enclosures. I’d be interested in a [url=http://www.lakka.tv/<]Lakka port[/url<] down the road, that's for sure.

      • Magic Hate Ball
      • 3 years ago

      I wonder if this one will handle N64 emulation better than the RaspberryPi 3 does.

        • derFunkenstein
        • 3 years ago

        The Pi3 does a pretty decent job already, but I’m sure a little more horsepower would touch up the rough edges. I’ve only dabbled with N64 on mine, though. I’m sure there are games that don’t run well.

          • evilpaul
          • 3 years ago

          Outside Mario 64 they basically all run between not great, badly, and completely unplayable.

            • derFunkenstein
            • 3 years ago

            ah, that and NFL Blitz is about all I’ve played, and they both seemed fine.

      • srg86
      • 3 years ago

      These little boards seem to live and die by their software support. The RPi is particularly good here. What is Rockchip’s support like?

      I actually think software support is more important than how good the hardware is.

        • derFunkenstein
        • 3 years ago

        This SoC uses stock ARM CPU and GPU cores, so support should be decent assuming a functional boot loader. Not sure about whatever networking chips it uses.

          • srg86
          • 3 years ago

          There’s no such thing as stock when it comes to these SoCs it seems, usually they get a few special kernels and then stagnate into nothing. I’m not sure if the ARM core or the GPU core being stock is enough.

            • derFunkenstein
            • 3 years ago

            That’s probably a good point. I don’t know enough about the technical workings to really say anything authoritative.

            • bhtooefr
            • 3 years ago

            It’s because there’s not been a standardized [i<]system[/i<] architecture for ARM platforms. If you're booting on a PC, you know what the memory map of your most critical peripherals is the moment the BIOS (or UEFI) hands control over to you, because all PCs have those most critical peripherals in the same place. You know that (on a BIOS system) if you call INT 13h, you'll be able to get sectors from the hard drive, if you need to load drivers for something. (And, quite a few other interrupt handlers that the BIOS gives you, letting you not worry about things until you're fully up and running.) If the PC's from approximately 1995 or later, you know you've got PCI, and that your BIOS has configuration tables telling you where your PCI devices are. If it's from the late 90s or later, you know you've got ACPI (which will tell you where a lot more things are). This holds true for literally every IBM PC compatible ever built, until approximately 2013, when UEFI-only devices without a Compatibility Support Module started coming out. And, those, the rules still hold, it's just that UEFI does those things for you instead of the BIOS. Let's compare to the situation on ARM. On ARM, U-Boot or whatever hands control over to you... and that's it, you need to know where everything is. And, each SoC vendor (and sometimes SoC model line) puts things in different locations, so you need to have been compiled for [i<]that SoC[/i<]. ARM has worked on this, with the Server Base specifications. One of the specifications is conceptually like the PC architecture documentation that Intel maintains for PCs, and specifies where things should be in a SoC, but it's optional for Server Base compliance. The second specification (Boot Specification) is for the firmware, and is, oversimplified, "read the UEFI 2.5 and ACPI 6.0 specs, make sure you comply with that". So, with Server Base Boot Specification, a kernel can use UEFI services while it's still coming up, and then use ACPI to figure out where things actually are so it can take control from UEFI. Worth noting that Windows 10 for ARM will require UEFI and ACPI, and apparently will boot on anything with it, based on tweets from people who have seen betas...

            • willmore
            • 3 years ago

            What you say is true at least before the device tree was added to Linux. Since then, the DT alows generic kernels to support specific hardware because uboot–which also supports DT–passes the DT file to the kernel when it starts it up. This lets the kernel know what hardware is where.

            I’ll have to add the “if UEFI is the answer, it must have been a very stupid question”.

          • Narishma
          • 3 years ago

          ARM GPU support on Linux isn’t that great.

        • Sgtkeebler
        • 3 years ago

        Rockchip is pretty nice, it is in the Asus Chromebook flip which is an amazing Chromebook.

Pin It on Pinterest

Share This