back to article Intel redesigns flawed Atom CPUs to stave off premature chip death

Intel finally has reworked its flawed Atom C2000 chips, which have been failing at a greater-than-expected rate for about a year and a half. On Wednesday, through an update to its chip errata [PDF], Intel revealed that its Atom C2000 chip family has a new C0 stepping, up from the previous B0, meaning at least some parts of the …

  1. Anonymous Coward
    Anonymous Coward

    How many manufacturers that got burned using the Atom C2000 will give Intel another shot?

    It's time for the author to get out there pounding the pavement asking...

    ----

    Love Intel, but I'll be avoiding any product that has Atom inside. (where ever possible)

    Atom is Intel's bastard step child, that's hidden in the closet when guests are over.

    1. Korev Silver badge

      How many manufacturers that got burned using the Atom C2000 will give Intel another shot?

      Probably more than you think. Remember "Pentium inside, can't divide" - Intel chips are ubiquitous these days. Moreover, moving off Atom to ARM, MIPS etc. would require at the minimum a recompile and in reality a lot more coding effort.

    2. Voland's right hand Silver badge

      More than you think.

      It is either giving it a try and having a part for 1:1 swaps or redesigning an entire piece of equipment.

    3. Anonymous Coward
      Anonymous Coward

      I started avoiding Atom, especially for NAS, after using an Atom netbook and seeing how damned slow it was, 720p wasn't playable on it! Even AMD embedded chips were better! The lowest Intel I use now is a parity RAM supporting Celeron.

      1. Alan Brown Silver badge

        The Atoms in question only have the name in common with old school ones.

        Whilst low power/speed compared to the high end products they have more cache and compute power than Xeons of 5-7 years ago (and would give a core i7-860 a run for its money whilst drawing 1/4 the power)

        Anaemic cacheless wonders of the early-mid 2000s these are not.

  2. Anonymous Coward
    Anonymous Coward

    Stepping != redesign

    It's an incremental improvement, often including bug-fixes. A redesign is a major work item, which this is not.

    1. diodesign (Written by Reg staff) Silver badge

      Re: anonymous

      Hi Brian!

      C.

    2. anonymous boring coward Silver badge

      Re: Stepping != redesign

      Apparently they had to remove some features. That's a re-design.

  3. John Smith 19 Gold badge
    Unhappy

    Let's see what this does to their earnings

    I don't know what the lead time on major chip purchases (and the processor is a major chip in this context) so maybe there will be no dip.

    But down the line people will be considering.

    1)How long Intel took to let anyone know

    2)How long it's taken to fix.

    3)How much trouble will a board re-design take accommodate their changed protocol

    4)How much trouble to go with ARM/MIPS/PowerPC/SPARC and rebuild their software.

    Maybe people will think Intel were OK in their response and it's too much hassle to change. Then again maybe not.

  4. Steve K

    BZZZT! "Reached Out"

    "Reached Out" - nooooooooooooo please, stop it.

    1. Anonymous Coward
      Anonymous Coward

      Re: BZZZT! "Reached Out"

      And......... cue The Four Tops

    2. anonymous boring coward Silver badge

      Re: BZZZT! "Reached Out"

      So humanitarian!

  5. Richard Tobin

    Clock Bus Clock

    You seem uncertain as to whether it's the LPC Clock Bus or the LPC Bus Clock. Presumably the latter, since the idea of a low-pin-count clock doesn't make much sense.

  6. Mage Silver badge

    Atom Cat?

    Has Intel lost the mojo?

  7. Version 1.0 Silver badge

    Out of Warranty

    They should be able to make money with this. Most devices sold have a one year warranty and if your Atom doesn't die for 18-24 months or longer, what are you going to do?

    Most people will just say feckit and buy a replacement device - it's the quickest way to get back up and running.

    1. theblackhand

      Re: Out of Warranty

      Not sure Intel will get out of it that easily...

      Most of the vendors have longer support warrenty programs for enterprise gear so HAVE to replace faulty equipment. i.e. Cisco set aside US$125million for "product replacement" (http://www.crn.com/news/networking/300083778/cisco-cfo-doesnt-anticipate-any-massive-revenue-impact-from-faulty-clock-components-company-sets-aside-125m-for-product-replacements.htm)

    2. Rol

      Re: Out of Warranty

      Throughout the EU, and yes that means Blighty as well, high ticket electrical items (more expensive than a kettle) are covered for 5 years.

      A TV with one years warranty does not need a further, "pay through the nose" extended warranty, as despite what the manufacturer and retailer are bleating, it has FIVE years cover against manufacturing errors.

      So the Atom being fundamentally flawed is sufficient for any EU citizen to claim a refund / repair if that component fails within five years.

    3. Alan Brown Silver badge

      Re: Out of Warranty

      "Most devices sold have a one year warranty and if your Atom doesn't die for 18-24 months or longer, what are you going to do?"

      These buggers are soldered-on devices - and frequently sold in systems along with 3-5 year support.

      System builders using stuff like Supermicro boards in their kit (or anyone else selling extended support options) are taking a bath on costs. This is the kind of event which puts smaller businesses out of action.

      On the other hand, you could do what HPE have just done to us and retroactively define your warranty (their "limited lifetime warranty" on LTO media is now only "3 years", despite being sold as "30 year archival or 260 complete backups")

  8. Anonymous Coward
    Anonymous Coward

    Intel design defects known as errataerrata ©

    "The products described may contain design defects or errors known as errata which may cause the product to deviate from published specifications." ref

    1. Anonymous Coward
      Anonymous Coward

      Re: Intel design defects known as errataerrata ©

      That's interesting, because it's true. The thing about this problem though is that it is not fixable through a firmware or microcode update. It bricks the Atom. Electronic components are finicky things. I designed a TTL circuit with three chips (hooray for the 555!) back in the 1980s to do telephone circuit automated testing. When I couldn't get a part from one manufacturer I found the competitor's replacement component to be slightly to wildly different in operation, and then I had to make up for that with a slight circuit redesign. Similarly, with the B0 C2000 Atoms, there must be no way to externally provide for the loss of the clock signal/function at that pin, nor other software method fixes. Only the C0 redesign cures it. That's pretty serious, but in the end I don't think it will make Intel lose it's place to AMD or ARM, but it can't help. I generally avoid the low-end CPUs, if there's not a great reason to use it and price/performance is not up to snuff. But with these, they are embedded in routers and other support gear, so everyone is potentially a victim. The vendors will decide if this is an "actionable offense" and do what they need to. Me? I've had a $35 ARM "server" in operation 24/7 for the past two years and am pretty impressed with it. Intel should make the Edison gear more like a rPi. There is just something about the rPi kit that makes it dead simple to use, and super easy to put into the shopping basket. Try harder, Intel!

      1. Alan Brown Silver badge

        Re: Intel design defects known as errataerrata ©

        "it is not fixable through a firmware or microcode update. It bricks the Atom. "

        And to rub salt into the wound - these are not socketed CPUs.

        It's notable that Supermicro no longer feature _ANY_ Atom systems in their lineup.

      2. Flocke Kroes Silver badge

        Re: Intel Pi

        Not going to happen any time soon.

        Intel got to the top the same way every CPU maker before them: build something cheap, sell huge numbers to divide the design costs of version N+1 down to insignificant. Their challenge for the past decade has been to stop cheap ARM CPUs doing the same thing to them. Intel have the most expensive manufacturing process in the world. ARMs get built with a manufacturing process that is one or two generations old and still give excellent performance per watt. When Intel make small CPUs it reduces their output of high margin big CPUs. If Intel design too small, people will buy cheaper ARMs instead. If they design too big, people will buy Atoms instead of Core. The Intel solution is a compromise with extra hurdles to buy and use Atoms.

        The only reason Intel would make something like a Pi is if they lost effective monopoly pricing on Core and Xeons. That will happen when you can slot your phone into the shell of a laptop (keyboard+trackpad+display but no CPU). The desktop equivalent would to to plug your phone into the USB C video/charge port on you TV which presents a bluetooth keyboard and mouse as USB devices to your phone. Add an ARM alternative to Xeon and Intel's high end margins would be cut to the point where making cheap CPUs does no additional damage to their revenue.

  9. PNGuinn
    FAIL

    So what exactly have they done inside and what will be the consequences?

    Has anyone any idea what exactly have Intel done here? It sounds like a horrible frig.

    It sure looks like some kind of cheapest / least effort fix. I wonder what the time / cost implications were to doing it properly?

    Which raises the question "What's the expected time to failure now, Intel?"

    And, assuming the brute isn't soldered in, "May I have my replacement chip now, please?"

    In reality, I presume in almost all cases it will be, so it will be down to the device manufacturers.

    So its not likely to be "We'd like n new chips, please." Rather "ARRRAGH - you barstewards - you've borked the pinouts. We'll need n new boards as well"

    I suspect Intel'll have to cut some *very* good deals to keep their customers sweet, let alone keep 'em at all long term.

    Diddums.

    1. Down not across

      Re: So what exactly have they done inside and what will be the consequences?

      So its not likely to be "We'd like n new chips, please." Rather "ARRRAGH - you barstewards - you've borked the pinouts. We'll need n new boards as well"

      I read it more as being the case of "we'd like new chips please", unless you were using the LPC pins as GPIO in which case you would need to do a redesign given you'll be needing additional logic for the functionaliity you needed the GPIO for.

      1. anonymous boring coward Silver badge

        Re: So what exactly have they done inside and what will be the consequences?

        I read it more as being the case of "we'd like new chips please", unless you were using the LPC pins as GPIO in which case you would need to do a redesign given you'll be needing additional logic for the functionaliity you needed the GPIO for.

        It's quite odd that they couldn't even fix the bug so that the processor works as originally designed.

  10. Anonymous Coward
    Anonymous Coward

    Chips and boards

    Anyone know what proportion of the affected products are built around bought-in boards rather than being designed around the chip inside?

  11. MT Field

    One trick pony

    Big powerful processors.

  12. atalba

    What is the expected failure rate of this processor over 18 months? Is it larger than normal, or highly anticipated for all to fail eventually, with 18 months being the expected product life?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like