Comments Locked

12 Comments

Back to Article

  • ZeDestructor - Friday, October 3, 2014 - link

    The big chip with 6x SATA ports could make for an interesting low-power ZFS box... It even has ECC!
  • Drizzt321 - Friday, October 3, 2014 - link

    Yes indeed, plus a large number of PCIe slots available for more HBA cards. Or just for straight PCIe SSD card(s). Plus 2 DDR3 controllers (which is how many slots?), which means probably a decent amount of max memory.

    Although the part that it's muxed with the 4x1Gb Ethernet ports is concerning, but just don't use those ports since it appears there's a 1x10Gb and a 1x1Gb that aren't hooked up to the mux.

    The other thing the PCIe slots allows is Thunderbolt, or for more enterprise settings Inifiband or the like.
  • tuxRoller - Sunday, October 5, 2014 - link

    Maybe, but only if it has neon(which I think all armv8 do) and zfs makes use of it. Otherwise, you'd be doing compression and parity calcs on a fairly slow CPU.
  • jospoortvliet - Tuesday, October 7, 2014 - link

    Neon is an intricate part of ARMv8 so it sure is there. No idea if ZFS uses it but why not...
  • ZeDestructor - Friday, October 3, 2014 - link

    Ganesh: There's an error in your table: the APM887208-H1 has 2 USB3.0 host, not 1, according to the diagram.
  • ganeshts - Friday, October 3, 2014 - link

    I see what you are saying, but APM's datasheet is confusing in itself. The text documentation only indicates one USB 3.0 host with integrated PHY, but the block diagram indicates two USB 3.0 ports.
  • milli - Friday, October 3, 2014 - link

    I guess the diagrams were made by different people or something.
    I think both have one USB 3.0 controller which offers two ports.
  • ganeshts - Friday, October 3, 2014 - link

    I got official confirmation from APM that the block diagram is right, but the text in their datasheet was wrong. I have fixed the table.
  • azazel1024 - Friday, October 3, 2014 - link

    What is odd to me is that Helix2 is the newer/smaller node, but much "smaller" chip in terms of core count and features, which means that Helix1 is probably a HUGE leap in power consumption being a full node larger and being a much more feature rich chip.
  • dave1231 - Monday, October 6, 2014 - link

    ARM could become the biggest company in the world if they continue like this. What a model. Even the LITTLE CPUS can be used in cheaper smartphones like the Desire 810. What can Intel do with all its money without licensing?
  • TadzioPazur - Tuesday, October 14, 2014 - link

    Write-through L1 D cache, that's gonna hurt. Unless it has either high memory parallelism/load-store OOO, or hyperthreading of a sort (which it does not). This gets me thinking which segments this product is for, as low-performance cores with ECC memory (is it mandatory?) do not match up in my book. Probably frontend (webpage generation), cheap cloud nodes (if this things has required features, like memort/IO virtualization) or storage servers (someone already proposed this in the comments above).
    I guess we'll see when actual products are announced.
  • Krysto - Friday, October 31, 2014 - link

    If APM wants to take the lead hard and fast, it should put Helix 3 on 16nm FinFET as soon as it can (hopefully early 2016 at most - late 2015 would be ideal).

Log in

Don't have an account? Sign up now