Hardkernel ODROID-XU2 Development Board with Exynos 5420 Coming Up in Q1 2014

Insignal recently announced that Arndale Octa with an Exynos 5420 was available for pre-order with shipping expected in December 2013, making it the first company providing a low cost big.LITTLE development board. But there will be some competition as Hardkernel will launch ODROID-XU2 also based on Samsung Exynos 5420 octa-core processor according to a tweet by SamsungExynos official twitter account. The board is also briefly mentioned in an article in Samsung website as it was showcased at ARM Techcon 2013.

ODROID-XU2 (Source: ARMFlix Video)
ODROID-XU2 (Source: ARMFlix Video @ 1:41)

That’s actually all the information I have. There’s a brochure on ARMFlix’s video, but it’s unreadable. However, we can derive (guess) some of the specifications based on the picture above and ODROID-XU specs:

  • SoC – Samsung Exynos 5420 octa core SoC with 4x Cortex A15 cores up to 1.8 GHz, 4x Cortex A7 cores up to 1.3 GHz in big.LITTLE configuration, and ARM Mali T-628 MP6 GPU
  • System Memory – N/A, but likely to be 2 GB (like ODROID-XU) or 3 GB RAM (like Arndale Octa).
  • Storage – micro SD slot, eMMC 4.5 connector
  • Video Output – micro HDMI + MPI LCD interface and Touch screen I2C ports (At the back of the board, not seen on pic) + DisplayPort (Bottom left)
  • Audio – micro HDMI, headphone jack, audio codec
  • Connectivity – 10/100M Ethernet via USB 2.0
  • USB – 1x USB 3.0 Host, 1x USB 3.0 OTG, and 4x USB 2.0 Host
  • Expansion Header – 30 pins (GPIO/IRQ/SPI/ADC)
  • Misc – PWM connector for Cooler, backup battery connector, serial console (UART) connector, and PMIC.

Price is unknown at this time. For reference ODROID-XU is available for $169 + shipping, so this new board is likely to be slightly more expensive once it becomes available in Q1 2014.

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard

23 Replies to “Hardkernel ODROID-XU2 Development Board with Exynos 5420 Coming Up in Q1 2014”

  1. Not applicable to this post, but I have no idea how to otherwise get a message to cnxsoft.
    Apparently the coupon code AFFM7 for focalprice is no longer valid. Guess it should be removed from the page.

  2. @onebir. The Intel stuff has much more performance per dollar. The main point of these boards is power and size: Odroid-U2 for example is roughly 5x5x5cm and you can use it without any active cooler. A perfect fit for a small home server that stays on 24/7 and sips a total of one or two watts on idle.

  3. Ordering as soon as it’s available. Still, we need those 64-bit Exynos 6 CPUs (14nm FinFET) with A53/A57. It’s since that Samsung already announced 64-bit Exynos for 2014 and there will be two version. One with official ARM cores (A53/A57) then Samsung will followup with custom built cores (Qualcomm-like). They also talked about 10nm already. I hope dreams come true and we get 14nm FinFET 64-bit Exynos 6 early in 2014 and dev boards in H2 probably :/

  4. @onebir
    For applications that support multiple core, the performance should be similar. Phonorix tested ODROID-XU (Exynos 5410) against a PC powered by Intel Atom D525, and the Exynos 5410 easily beats the Atom processor for all tests. http://www.phoronix.com/scan.php?page=article&item=arm_odroidxu_octa&num=2

    Intel Core i3 2100T is faster than Atom D525, so I assume it would probably match Exynos 5410 is these particular benchmarks. However for single thread performance, Intel Core i3 should still have a clear lead over the Cortex A15.

    Beside performance, the most important point is that the Intel Box is just ready to use for end users, whereas Hardkernel and Insignal boards are mainly for development, and software may not work as well as on the product you linked to. I’ve seen complains for both Arndale and ODROID boards. It seems that if you’re just looking for something with a mostly working Ubuntu and Android, with some community support, Hardkernel ODROID boards may be better. But if you want to work with the latest mainline kernel, and accept there may not be that much support, you’d better go with Arndale. That’s my understanding anyway based on users’ feedback I’ve read on the web. If you’ve used the boards, feel free to disagree, and/or correct me.

  5. @Joe Bucks
    This Core i7 CPU is not really a performance beast, and is equivalent to a Core2 Duo E6600 @ 2.40GHz

    So still the performance is better than Atom based system, so the price looks good ($200 with shipping by courier). There’s another seller with a better rating that sells a similar system with more RAM (4GB) but no storage for $180 with shipping by HK post. http://www.aliexpress.com/item/PC-Desktop-Barebone-Mini-Barebone-PCs-with-intel-dual-core-i7-U620-1-07Ghz-8-usb/1355766588.html

  6. Why are these boards coming out so late?

    Does it really take such a long time to do them or does Samsung only supply their “Mercedes” chips to this hobbyist niche, once they no longer have any use for them in their own smartphones and tablets?

    With 64-Bit ARM chips supposed to be just around the corner these extened 32-bit ARM15 seem to be dead unless there was a really good price argument, yet I can’t see Samsung addressing the RK3188 market with Exynos 5.

    So who will really want these board now? (I know I would have stood in line a year ago)

    The developers mostly will have moved on to 64-bit, the consumer device space would seem content to jump from 28nm ARM9 and ARM7 directly to ARM53 or 57 and the hobbyist will most likely demand more flexibility.

    I find that I’d want SATA and Gbit Ethernet onboard as well as more RAM (perhaps even a SODIMM slot), so I can actually start using Xen or KVM on ARM to run Ubuntu and Android on the same device (Android to serve the TV screen/projector, Ubuntu to run the fileserver/proxy/FW/Plex server infrastructure for the phones and tablets).

    But SATA support got lost again on 2nd generation Exynos 5.

    I find it irritating that AMD and Intel are doing better at coming down (e.g. Kabini) than ARM/Samsung at going up.

    I can’t fault the ARM camp for lack of vision, because that’s all there and announced.
    But the ability or will to execute on Eynos 5 by Samsung seems a completely different thing.
    Perhaps chosing Samsung as a launch partner for ARM15 was a mistake.

    BTW: Don’t believe these photos! You won’t be able to run this board without at least a very large passive heat sink.

  7. @Joe Bucks
    Sorry I have not explanation. I don’t follow Intel that much, but apparently it’s supposed to work as follows:
    * Core i3 – low-level performance
    * Core i5 – mid-range performance
    * Core i7 – high-end performance

    However that i7 processor has lower performance than most of the Core i3 processors.

  8. @Thomas Hoberg
    The way I understand it, there is a time period that must pass after the release of a new SoC by Samsung before any company that is planning on releasing a device using that SoC is allowed to release their product (something like 6 months). It’s possible that ARM might be twisting Samsung’s arm (lolpun!) to allow Insignal and Hardkernel to release their 5420 board before that time period is up given the supposed CCI-400 bug of the 5410 SoC (even if the CCI-400 bug doesn’t keep the 5410 from running in true HMP mode, ARM would rather the dev boards on the market have their Mali-T628-MP6 GPU represented than the ImgTec PVR544-MP3).

    The ARM Cortex-A53/A57 chips are purpose-built for low-power server hardware for the increasingly tightly-packed datacenter. I don’t think the A53/A57 are really aimed at mobile devices. If this is true, then you won’t find them in dev boards… only in the rather expensive enterprise market, where it is easy to gouge a hefty premium out of the end user for the joy of the newest low-power server equipment (which saves you not only in power consumption dollars, but in cooling costs too). The hobbyist cannot afford to pay for a $5k server, and I doubt very much if anyone would be willing to jump up even to the comparitively modest cost of a decent desktop x86 machine ($600-$1200).

    There is absolutely no need for 64-bit addressing in any mobile chip. Where are you going to find a phone with more than 4GB of RAM? Come to think of it, I don’t really know of any phones with more than 2GB (not to say that one doesn’t exist… just that I haven’t heard of it). If you don’t have more than 4GB of addressable memory, then there is no reason to need a 64-bit-capable CPU.

    The reason the Exynos SoC isn’t built to be a desktop system’s CPU is because it was built to be the application processor in a cell phone. To be fair, the purpose for these development boards is not to be a desktop system, either. They are meant to be an open platform on which to develop software for phones, mobile devices, and embedded systems. When tested against those purposes, the Exynos 5 is actually far, far more powerful than it needs to be, which is why many people have been finding themselves wondering why gigabit ethernet, SATA2/3, and DIMM/SO-DIMM slots weren’t included. While I find myself wishing for the same things myself, I also realize that adding those things doesn’t serve the market very well — you would expect to add something like $30 for the gigabit ethernet capability, and the SO-DIMM slot(s) and SATA bus aren’t supported by the SoC.

    Once you realize what the ODROID and Arndale boards are built for, you really have to be amazed at how powerful they are. The Exynos 5 SoC was never even meant to run Ubuntu, but thanks to the work and dedication of an awesome team of hardware and software engineers and a ridiculously overpowered mobile CPU we find ourselves attempting to replace our home desktop and server systems with these boards. While people really have no business buying these dev boards if they aren’t Android, embedded Linux, or hardware developers, I don’t see the demand for them slowing down. This is especially true for Hardkernel, thanks to how big their community has grown and how helpful HK’s developers and community are.

  9. @mjstanton
    Just a comment about 64-bit… Initially, I thought the same, that it would be useless in smartphone for now because device are not using 4GB or more RAM, but there’s more to the move from ARMv7 to ARMv8 (increase in the number of general purpose registers, doubles the number of FP/NEON registers, hardware AES,…) than meets the eye, as Anandtech found out in their review of the iPhone5s: http://www.anandtech.com/show/7335/the-iphone-5s-review/4

  10. @cnxsoft
    Exactly! People don’t think that it’s a completely new ISA, double registers, double precision support in NEON, etc. It’s almost like a new chip, but still compatible in AArch32 mode for ARMv7. Currently we have 3GB RAM max in smartphones, but 4GB also coming.

  11. @mjstanton
    I don’t believe the A53/57 are purpose built for the data center. They are building blocks, which can be used individually or together depending on your use case.

    For starters, there wouldn’t be an A53 companion in the datacenter incarnation of ARMv8, because in a Moonshot enclosure full with A57s you’d just turn off the ones you don’t need at the moment. At the same time you see the Altera FPGAs produced on Intels 14nm process using ARM53 exclusively, because on these chips they aren’t used for compute workloads, but for administration/programming of the FPGAs.

    This BIG/little combination though is really to support the mobile use case, which oscillates between almost desktop intensity and the compute power you expect from a toaster.

    It also shows demonstrates just how cheap CPU transistors are these days in terms of silicion real estate.

    ARM tries to be as generic as possible in general and 64-bit simply is the next logical growth step for ARM, as it was for all other general purpose architectures.

    I give Apple credit where it is due (grudgingly) and their choice and move to go to 64-bit is mostly about simplifying there future software ecosystem: Everything will be 64-bit in their future. And there is some speculation that also everything will be ARM 😉

    Note that they stayed with dual-core for the mobile devices: Just like nobody really needs 64-bit on a mobile device today, nobody really needs quad-core CPUs either. Yet even the lowliest smartphones come with quadcore CPUs today, simply because…
    – it sounds good
    – it the incremental cost to dual core is peanuts
    I’ve seen figures like 15% of a SoC’s real-estate being consumed by the quad core CPU, so slicing off two cores means you save roughly 8% in die area.
    Going to 64-bit costs perhaps a 20% premium on the CPU die area, add another 10% for the difference in instruction sets of ARM32 vs. ARM64 and a 64-bit dual core CPU is a lot smaller than a 32-bit quad (but still just a minor portion of the SoC).

    The bigger difference today would be the licensing cost to ARM, where any first A57 (and Mali T6xx/T7xx) costs premium and an A9 is cheap. Since additional cores of the same type in a chip typically cost peanuts also in licensing, that explains why these “cut&paste designs” are so popular in cheap SoCs.
    With A9 Quads and A7 Octas (yes, these are being announced) occupying perhaps a similar silicon real-estate and licensing space at the low end, the A15 and A12 designs are in a niche that is closing extremely fast.

    All these A7 guys I see jumping directly to A53 using 4 or 8 cores just to advertise “64-bit 8 core!!!!” and A57/53 designs will capture the high-end.

    I simply can’t see Samsung selling Exynos 5 in high volumes any more (at the prices they’d want) and I wouldn’t be astonished to see them dump the whole line rather fast.

    And that’s what takes me back to the beginning and my very first question: Since developers go where the future and the money is, this last round of high-end 32-bit devices is so close to becoming obsolete in terms of *architecture* that they don’t seem to make sense any more as developer boards.

    In terms of *functionality* ARM15 would be adequate for many use cases for many years, especially since it *does* support quite a bit more RAM addressing capability, a full TERABYTE in fact (which is exactly why I always wanted to have a SO-DIMM socket even on my Arndale board so I can play around with 8GB or more using Xen/KVM).

    For a *software developer* that board doesn’t make much sense beyond the ARM-15 boards already available, because the future is 64-bit and will support 32-bit development just was well.

    For a *device developer* I don’t see Exynos 5 hitting an attractive price point below Eynos 6 or close enough to an RK3188 and for sufficient time to make a design worthwhile. Very soon A53s quads will occupy that space and those people smart enough to realize that the A15 would be faster will just go for the A57. Qualcom Snapdragon rules the LTE space and the pure WiFi niche is ruled by price, which so far seems to exclude Eynos.

    So the only way these boards would make any sense today for *hobbyist* is that Samsung will have everyone wait again, before thay’d allow an ODROID to come out for the 64-bit Exynos they have already pretty much announced.

    As to leaving out SATA and Gbit Ethernet: I simply don’t understand that in terms of chip area. I don’t know what it would cost in licensing and testing, but SATA has been there in the dual core Eynos 5 and works just fine on my Arndale, even with SATA port multipliers. It allows Arndale to work both as a credible full HD desktop, including light gaming and as a 10 Terabyte home media server, which just happens to use 2Watts in active idle (not the sort of hibernate you’d have to do with x86 to get to that power footprint) with all disks spun down.

    For the mobile phone SATA benefits are dubious but if it avoids doing a distinct chip for Chromebooks or STB why not put it in or offer it as a variant?

    Samsung doesn’t seem to have a clue so they put it in (40nm Exynos 4) the take it out (32nm Exynos 4), they put it in (32nm Exynos 5) and they take it out (28nm Exynos 5).

  12. @davidlt

    Why to order this, if its already outdated? (64-bit boards coming out sooner than Q1 2014)
    There is already 5420 board, no need to wait if you really want to buy that EOL

Leave a Reply

Your email address will not be published. Required fields are marked *

Khadas VIM4 SBC
Khadas VIM4 SBC