Stories
Slash Boxes
Comments

SoylentNews is people

posted by cmn32480 on Saturday December 05 2015, @06:28PM   Printer-friendly
from the same-designer-as-the-iPhone-6 dept.

Found this at HotHardware. From the article:

It's been discovered that some third-party heat sinks can physically damage Intel's new Skylake CPUs, along with the pins in the accompanying motherboard socket. The problem has prompted at least one cooler maker to change the design of its Socket 1151 heat sinks and it wouldn't be surprising if others soon followed suit.

The apparent issue is the substrate Intel used for its Skylake chips. A close-up shot of a Skylake CPU sitting side-by-side with a Broadwell processor shows that the substrate is noticeably thinner on Skylake, and thus prone to bending from the force that some third-party heat sinks exert. It also poses a problem for the tiny pins in the socket area of Skylake motherboards.

Sounds like something to be careful of when building that new rig. Has anyone experienced the issue?

takyon:

Update - 3:08PM: This just in from Intel...

"The design specifications and guidelines for the 6th Gen Intel Core processor using the LGA 1151 socket are unchanged from previous generations and are available for partners and 3rd party manufacturers. Intel can't comment on 3rd party designs or their adherence to the recommended design specifications. For questions about a specific cooling product we must defer to the manufacturer."

And so it would appear this is an OEM 3rd party manufacturer issue, rather than a generalized issue with the processor(s).


Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @06:41PM

    by Anonymous Coward on Saturday December 05 2015, @06:41PM (#272228)

    The most significant SIMD innovation of the decade will now be generally regarded as crap by association. AVX3 is crap! because Skylake is crap!

    • (Score: 3, Interesting) by takyon on Saturday December 05 2015, @07:27PM

      by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Saturday December 05 2015, @07:27PM (#272249) Journal

      Intel ‘Skylake’ processors for PCs will not support AVX-512 instructions [kitguru.net]

      I'm pretty sure only Xeon and Xeon Phi support that anyway. And those chips sure as hell won't be warping.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 2) by frojack on Saturday December 05 2015, @08:23PM

        by frojack (1554) on Saturday December 05 2015, @08:23PM (#272261) Journal

        Can someone explain why I might give a shit about AVX3?

        I've not had it all along, so I'm unlikely to miss it, because I don't even know what it is.

        --
        No, you are mistaken. I've always had this sig.
        • (Score: 3, Informative) by gman003 on Saturday December 05 2015, @09:00PM

          by gman003 (4155) on Saturday December 05 2015, @09:00PM (#272268)

          AVX is a series of SIMD instruction set extensions, similar to SSE or MMX or 3DNow.

          AVX and AVX2 generally just extended SSE from 128-bit to 256-bit operation. AVX was introduced with Sandy Bridge ("second generation i7"), AVX2 was introduced with Haswell ("fourth generation i7"). They're fairly well-supported by compilers now, so as long as you're building your own binaries you've probably gotten a performance boost on anything that does a lot of parallel math. I believe they are also supported on certain AMD chips, but I haven't looked into all the details because it looks like a messy subject.

          AVX-512 further extends it to 512-bit operation. This is the width of the parallelism, not the actual data, mind you - SSE's 128-bit SIMD was four 32-bit or two 64-bit operands, AVX's 256-bit is eight 32-bit or four 64-bit operands, and so AVX-512 works on sixteen 32-bit or eight 64-bit operands. This was designed for the Xeon Phi coprocessor - a PCIe card used mainly in supercomputers, where crunching that much data is commonplace. It was intended to be launched on Skylake ("sixth generation") Xeons, not sure if that's happened yet. There are no consumer processors on Intel's roadmap with AVX-512 support, and compiler support isn't widespread AFAIK so unless you hand-write code for AVX-512 you might not see as much performance improvement.

          AVX3 doesn't exist yet. That original AC has no idea what he's talking about.

          • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @10:53PM

            by Anonymous Coward on Saturday December 05 2015, @10:53PM (#272290)

            AVX3 and AVX-512 are the same thing, and GCC has been capable of automatically generating AVX-512 instructions since 2014.

        • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @11:08PM

          by Anonymous Coward on Saturday December 05 2015, @11:08PM (#272297)

          mod up the willfully fucking ignorant shithead

          way to go bro

  • (Score: 3, Interesting) by MrGuy on Saturday December 05 2015, @06:47PM

    by MrGuy (1007) on Saturday December 05 2015, @06:47PM (#272229)

    This just in from Intel...
    "The design specifications and guidelines for...using the LGA 1151 socket are unchanged... Intel can't comment on 3rd party designs...For questions about a specific cooling product we must defer to the manufacturer. And so it would appear this is an OEM 3rd party manufacturer issue, rather than a generalized issue with the processor(s).

    There's an issue with some heatsinks working with a particular processor. The heatsink manufacturer blames the processor manufacturer. Then the processor manufacturer comes back and blames the heatsink manufacturer.

    I'm not sure how your conclusion "therefore, it's the heatsink manufacturer's fault, and not the processor" follows from those facts.

    • (Score: 2) by moondoctor on Saturday December 05 2015, @07:03PM

      by moondoctor (2963) on Saturday December 05 2015, @07:03PM (#272236)

      What they are saying is that the coolers exceeded specs. Loosely: The old ones could handle more pressure, even though the pressure was not within spec. The new ones buckle under an over-spec load.

      After this statement it is now about whether the cooler folks were selling units that put pressure beyond the spec.

      We'll see...

      • (Score: 3, Informative) by frojack on Saturday December 05 2015, @07:58PM

        by frojack (1554) on Saturday December 05 2015, @07:58PM (#272255) Journal

        The old WHAT could handle the pressure?

        The problem is that the the coolers are deforming (and possibly breaking) the CPU chips.

        A close-up shot of a Skylake CPU sitting side-by-side with a Broadwell processor shows that the substrate is noticeably thinner on Skylake,

        Skylake a new chip. There are NO OLD coolers that should be used, the new chipset requires a new cooler. End of story.
        Manufacturers are sending out their old Broadwell coolers without a glance at the specs.

        So far it only looks like the Scythe brand [extremetech.com] of coolers are at fault.

        --
        No, you are mistaken. I've always had this sig.
        • (Score: 2) by maxwell demon on Saturday December 05 2015, @08:43PM

          by maxwell demon (1608) on Saturday December 05 2015, @08:43PM (#272265) Journal

          The heat sink makers relied on undocumented behaviour (the processor being able to stand a larger force than specified) which changed for the newer processors. Any heat sink for the old processor that was manufactured to spec should work for the new processor as well.

          Let me make a car analogy.

          Intel made bridges and told people that only cars up to 5 tonnes weight should drive over them. But some car manufacturers noticed that they could drive 7 tonnes cars over those bridges without them breaking, and sold such cars.

          Then Intel started to make new bridges that would break when you drive 7 tonnes cars over them, however 5 tonnes cars still can drive over them without problems. Therefore all cars which officially were able to cross the old bridges can also cross the new bridges. However the cars that are heavier than allowed will break the new bridges even though they happened not to break the old bridges.

          --
          The Tao of math: The numbers you can count are not the real numbers.
          • (Score: 2) by frojack on Saturday December 05 2015, @09:28PM

            by frojack (1554) on Saturday December 05 2015, @09:28PM (#272275) Journal

            Any heat sink for the old processor that was manufactured to spec should work for the new processor as well.

            No, it shouldn't. They are different chips, with a different thickness substrate, for a different market segment (thin tablets), and they had DIFFERENT SPECS WERE PUBLISHED by Intel. Some 3rd party suppliers noticed the old heat sinks sorta fit, and decided to pawn off old equipment for the new processors.

            You must be new to computers. Here's a clue: Things change rapidly in the computer industry, mkay? Remember that fact. It will serve you well later in life after you are out of Junior High School. What works today, can not be expected to work tomorrow with different motherboards, different chips.

            Let me make a car analogy.

            Sorry, NO. Cars are not involved. These are not the bridges you were looking for,

            --
            No, you are mistaken. I've always had this sig.
            • (Score: 3, Insightful) by maxwell demon on Saturday December 05 2015, @09:46PM

              by maxwell demon (1608) on Saturday December 05 2015, @09:46PM (#272277) Journal

              and they had DIFFERENT SPECS WERE PUBLISHED by Intel.

              FTFS (emphasis by me):

              The design specifications and guidelines for the 6th Gen Intel Core processor using the LGA 1151 socket are unchanged from previous generations and are available for partners and 3rd party manufacturers.

              Note that, according to the summary, that's a direct quote by Intel. Now who knows better about Intel specs, Intel or frojack? Well, hard call! ;-)

              You must be new to computers.

              No. But I'm able to read. That's a big advantage; you should try it, too.

              --
              The Tao of math: The numbers you can count are not the real numbers.
    • (Score: 2) by frojack on Saturday December 05 2015, @07:07PM

      by frojack (1554) on Saturday December 05 2015, @07:07PM (#272238) Journal

      Intel can't comment on 3rd party designs or their adherence to the recommended design specifications.

      And why not?

      If there is a brand known to work improperly, name-em and shame-em.

      Thanks a lot Intel, for dodging the issue here and leaving at the mercy of shoddy manufacturers. I bet similar reluctance won't be seen when we have to buy a new CPU because some lame fan broke the corner off of our processor. (Yes I have seen this happen in the past}.

         

      --
      No, you are mistaken. I've always had this sig.
      • (Score: 2) by Hairyfeet on Sunday December 06 2015, @12:38AM

        by Hairyfeet (75) <{bassbeast1968} {at} {gmail.com}> on Sunday December 06 2015, @12:38AM (#272311) Journal

        Nothing "shoddy" here except Intel, unless you are actually gonna buy a classic "who ya gonna believe, me or your lying eyes" from Intel. Look at the pics yourself, see the substrate? See how much thinner it is? Yet Intel claims that even though they cut the substrate in half that its no different than the previous chips specs..yeah, and pull this leg it plays jingle bells.

        --
        ACs are never seen so don't bother. Always ready to show SJWs for the racists they are.
      • (Score: 1) by tftp on Sunday December 06 2015, @12:41AM

        by tftp (806) on Sunday December 06 2015, @12:41AM (#272313) Homepage

        And why not?

        Lawsuits, of course.

    • (Score: 2) by sjames on Saturday December 05 2015, @08:23PM

      by sjames (2882) on Saturday December 05 2015, @08:23PM (#272260) Journal

      The specs from Intel are unchanged. There were some out of spec heatsinks that just happened to work OK with older CPUS. The new CPU is less forgiving of out of spec heatsinks but works fine with old heatsinks that are in spec.

      So it's the 3rd party's fault for having a heatsink that was never in spec.

      • (Score: 3, Informative) by MrGuy on Saturday December 05 2015, @11:04PM

        by MrGuy (1007) on Saturday December 05 2015, @11:04PM (#272295)

        The specs from Intel are unchanged

        Assertion supported solely by Intel's statement.

        There were some out of spec heatsinks that just happened to work OK with older CPUS.

        Unsupported assertion. Where is the evidence that the working heatsinks were out of spec?

        The new CPU is less forgiving of out of spec heatsinks but works fine with old heatsinks that are in spec.

        Unsupported assertion. Where is the evidence the failing sinks are out of spec? Where is the evidence that the CPU's DO work with in-spec heatsinks.

        So it's the 3rd party's fault for having a heatsink that was never in spec.

        Logical conclusion that follows if you accept these assertions as factual.

        I'm not saying you're not correct, just that I don't see evidence to support this argument presented here.

        Intel's statement (which supports the first assertion) is a statement from one of the parties in a dispute over who is to blame for an engineering issue claiming it's the other party's fault. I'm personally skeptical of taking their unsupported word for it. The second and third assertion are [citation needed] to me. They may be true, they may not.

        • (Score: 2) by sjames on Sunday December 06 2015, @06:49AM

          by sjames (2882) on Sunday December 06 2015, @06:49AM (#272409) Journal

          Intel's statement (which supports the first assertion) is a statement from one of the parties in a dispute over who is to blame for an engineering issue

          It is also a statement from the party that actually wrote the spec and published it. Although I don't have a copy, it would be so very trivial for any ticked off 3rd party to show the change if it existed that it would be the world's dumbest lie. They're smarter than that.

          It's a story that has been repeated so often in the industry that I don't find it at all difficult to believe, though I would welcome evidence to the contrary.

      • (Score: 0) by Anonymous Coward on Sunday December 06 2015, @12:21AM

        by Anonymous Coward on Sunday December 06 2015, @12:21AM (#272305)

        As far as I can tell, the specs are indeed unchanged, but they don't specify the disputed parameter.

        According to the article, Intel is saying that the 6th generation has the same specifications for mounting pressure as did earlier generations. I found the "Thermal Mechanical Design Guidelines" (TMDG) for the 4th generation of processors. [intel.com] This looks like it's meant to be the go-to document for designing coolers, but I found no mention of the "maximum mounting pressure" that the article says is specified. I didn't find any different TMDG document for newer processors that use the LGA1151 socket (they exist for those using the LGA2011 socket).

    • (Score: 0) by Anonymous Coward on Sunday December 06 2015, @09:33AM

      by Anonymous Coward on Sunday December 06 2015, @09:33AM (#272436)

      Ha!

  • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @06:52PM

    by Anonymous Coward on Saturday December 05 2015, @06:52PM (#272233)

    [quote]the 6th Gen Intel Core processor using the LGA 1151 socket are unchanged from previous generations[/quote]

    So intel is saying there was no change to specs, but their parts are wrapping and breaking.

    So, intel what did you change that makes these new chips out working specs? Worst pat is intel has the cash, to test heat sinks to detemining equality. 3rd party have the incentive to send to Intel to valuate. MY guess is Intel wants all other manufactures to die, so the increase profits again.

    • (Score: 2) by Nollij on Sunday December 06 2015, @06:27AM

      by Nollij (4559) on Sunday December 06 2015, @06:27AM (#272406)

      A lot of heatsinks break specs. I remember looking into this around LGA775. Almost every aftermarket unit on Newegg exceeded the weight limit.

      I bet the same thing is still true.

  • (Score: 2, Funny) by Anonymous Coward on Saturday December 05 2015, @06:53PM

    by Anonymous Coward on Saturday December 05 2015, @06:53PM (#272234)

    I'm so cool I don't need a heat sink when My Rig is cooled by the overwhelming coolness of my presence.

    • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @07:20PM

      by Anonymous Coward on Saturday December 05 2015, @07:20PM (#272242)

      you're alaskan for trouble

    • (Score: 2) by Runaway1956 on Saturday December 05 2015, @07:30PM

      by Runaway1956 (2926) Subscriber Badge on Saturday December 05 2015, @07:30PM (#272250) Journal

      So - I'll just wait until you are absent (not present and not president) to turn your computer on.

  • (Score: 0) by Anonymous Coward on Saturday December 05 2015, @07:48PM

    by Anonymous Coward on Saturday December 05 2015, @07:48PM (#272253)

    Like this one [dickssportinggoods.com] might not work so well with Skylake.

    • (Score: 2) by Username on Saturday December 05 2015, @08:58PM

      by Username (4557) on Saturday December 05 2015, @08:58PM (#272267)

      That one might actually work better. Fill it with mineral oil, and submerge the motherboard. Bonus points for a CRAY style waterfall.

  • (Score: 1, Informative) by Anonymous Coward on Saturday December 05 2015, @09:11PM

    by Anonymous Coward on Saturday December 05 2015, @09:11PM (#272274)

    If you look at a Broadwell and Skylake chip next to one another the Skylake has a significantly thinner substrate. So, no, it cannot handle as much load from a cooler.

    It is a sign of the times. These things are getting smaller and more delicate.

    [pic of the thinner waffer]-->http://images.anandtech.com/doci/9505/PCW-Thick.jpg

  • (Score: 2) by Gravis on Saturday December 05 2015, @11:41PM

    by Gravis (4596) on Saturday December 05 2015, @11:41PM (#272302)

    The design specifications and guidelines for the 6th Gen Intel Core processor using the LGA 1151 socket are unchanged from previous generations and are available for partners and 3rd party manufacturers.

    this seems like a collision of specifications vs reality. i suspect that the former generation could withstand far more pressure than the current generation of processors. as a result, people designed their heatsinks to work with reality instead of the specification. maybe they saved a few cents or something but as a result, the current generation is being crushed by heatsinks.

    the good news in all this is that AMD chips continue to be rock solid. :)

    • (Score: 0) by Anonymous Coward on Sunday December 06 2015, @12:44AM

      by Anonymous Coward on Sunday December 06 2015, @12:44AM (#272314)

      Even solid rock can be crushed. ;-)