Intel tackles NUC heat issues with fan speed tweak, SSD fix

This update is a bit overdue, but we’ve only recently had a chance to play with a final, production version of the Intel "Next Unit of Computing" system that we reviewed last year. Back then, our review sample suffered from an overheating problem that caused full-on system lockups during large file transfers over Wi-Fi. We worked with the folks at Intel to confirm that the problem affected all NUCs of that type, the red-topped "BY" model, and we played with a preliminary BIOS intended to resolve it. However, we were never able to confirm with certainty how the fix—likely a tweaked fan speed policy—would impact the system’s acoustic profile.

Fortunately, I’ve just tested a production NUC unit with the latest BIOS at its default settings, and I have generally positive news to report.

The BIOS, which I downloaded from Intel’s public website, sets a default minimum fan speed of 30%, regardless of which fan profile the user chooses. That’s a little higher than it was on our initial test system, and the system is perceptibly noisier, with a bit of a high-pitched whine when you put your ear right next to it. Still, the system is barely audible from two feet away. The end result isn’t as ideal, acoustically, as our first review unit, but it’s not bad, either.

Crucially, the system also withstands large file copies over Wi-Fi with the same SSD installed as in our initial test unit. We copied approximately 60GB of data over Wi-Fi in multiple tests of different file types, and although the air coming out of the back of the NUC got to be pretty warm, the system never locked up or exhibited any obvious quirks. It appears the NUC engineering folks have found a reasonably happy medium between heat management and acoustics.

I should note that the overheating problem we encountered involved not just the heat from the Wi-Fi card but also its effect on the closely situated mSATA SSD, which in our review system was a pre-release version of the Intel 525 Series SSD, a product formally released just yesterday. We learned in a conversation with some Intel reps at CES that the NUC lockups were in fact the result of two problems coming together: too much heat in the NUC chassis and a defective thermal throttling mechanism in pre-production Intel 525 SSDs. Intel tells us the throttling mechanism has been corrected in production drives, so in the future, any overheating should merely cause the drive to slow down, not to lock up and become completely unresponsive until the system’s power plug has been pulled.

We haven’t tested the production NUC with a production 525 Series SSD, but given that our throttling-challenged pre-production SSD doesn’t lock up with the new fan profile, we’re generally satisfied that Intel has ironed out the wrinkles in the NUC’s behavior. If you’re interested in buying one now that the coast is clear, they’re available at Amazon for $310 right now. Anyone who’s deploying a "BY" NUC in a particularly warm ambient environment may want to consider using a fairly aggressive cooling profile, though, just to be safe.

Meanwhile, Intel told us at CES that the NUC has "completely exceeded" its expectations for consumer adoption, so more systems are coming soon. Up next: NUCs that are both cheaper and more expensive, packing Celeron and Core i5 processors.

Comments closed
    • HisDivineOrder
    • 7 years ago

    I totally predicted they’d slow that SSD down to compensate.

    • Stuart21
    • 7 years ago

    INVENTORS – DO NOT TRUST INTEL
    I invented a CPU cooler – 3 times better than best – better than water. Intel have major CPU cooling problems – “Intel’s microprocessors were generating so much heat that they were melting” (iht.com) – try to talk to them – they send my communications to my competitor & will not talk to me.

    Winners of major ‘Corporate Social Responsibility’ awardS!!!

    Huh!!!!

    When did RICO get repealed?”

    INVENTORS – DO NOT TRUST INTEL!!!

    BTW, I have the evidence – my competitor gave it to me.

    BBTW, I am prepared to apologise to Intel if;

    • They can show that the actions were those of a single individual within the company, but acting outside corporate policy, and:

    • They gain redress on my behalf.

    Stuart Saunders,
    Intellectual Property Rightful Owners Action Group.
    reform@iproag.org

      • NeelyCam
      • 7 years ago

      Links, or it didn’t happen

        • Stuart21
        • 7 years ago

        Good luck.

        If it didn’t happen, why has Intel not taken the opportunity to correct the story, by providing their version of events – they’ve had lots of opportunity, I have been slagging them off wherever I can for nearly two years now. Google “INVENTORS – DO NOT TRUST INTEL “

          • chuckula
          • 7 years ago

          Hrm… I took your advice and all I saw were a bunch of cut & pasted random forum posts that have the exact same drivel you posted above. Hey TR Gods: How ’bout a nice dose of VITAMIN BAN-HAMMER to cure all that ails Stuart21 here.

            • Stuart21
            • 7 years ago

            If the TR gods would like to sign a secrecy agreement, I am happy to prove that all in my original post is true.

          • chuckula
          • 7 years ago

          [quote<]Good luck. If it didn't happen, why has Intel not taken the opportunity to correct the story, by providing their version of events [/quote<] Stuart21 SHOT KENNEDY! I know it's true because he hasn't posted HIS OWN VERSION OF THE EVENTS! He's had plenty of opportunity, so IT MUST BE TRUE!

            • Stuart21
            • 7 years ago

            I have posted my own version of events.

            I didn’t show them the proto.

            But you have seen all you are going to see.

            • Peldor
            • 7 years ago

            Another opportunity missed. Why the continual avoidance of the issue?!?

          • Deanjo
          • 7 years ago

          [quote<]If it didn't happen, why has Intel not taken the opportunity to correct the story, [/quote<] Because if they took the time to shoot down every nutjob post they wouldn't have time to innovate. BTW, as an accuser, the onus falls upon you to prove it, for which you have provided nothing other then an accusation.

            • Stuart21
            • 7 years ago

            And I have to prove it to whom?

            My warning is to inventors, they can take it or leave it. Done my bit.

            • Deanjo
            • 7 years ago

            Well first of all intel, secondly the courts if you wish to prove your claims to an authority that can do something about it. Otherwise you are just another nut job making grandiose claims that you can’t back up.

            • Stuart21
            • 7 years ago

            ‘Samatter with Intel? Cat got their tongue?

            Are you telling me I have to stop bullying poor little sensitive Intel?

            I don’t ‘want’ to prove anything to anybody. But I can.

            As I said, I just want to warn inventors.

          • NeelyCam
          • 7 years ago

          If your “competitor” got their hands on this “3 times better than best” cooling technology, I’m sure they have already brought it to the market… where can I buy it? A link to Amazon product listing would be greatly appreciated

            • Stuart21
            • 7 years ago

            Not stupid – neither Intel nor competitor have the design. Just a brief description.

            Prolly will never be manufactured – I also use the description ‘Patented by Smith & Wesson’.

    • Metonymy
    • 7 years ago

    Am I the only one who would happily see these things be 1/2″ to 1″ taller if just to accomodate a slightly larger fan so that they could be quieter and more … relaxed… without pushing the thermal envelope?

    I’m not being snide here: Are the applications for these such that it’s really not just a case of “how small can we make it” but that the they’d not sell as well slightly larger?

      • UberGerbil
      • 7 years ago

      Just as I thought with the very-similar Dell Zino, I’d like to see a mesh lid and a slow-spinning fan that covers almost the entire area below it. You might have an issue with RF but otherwise it would be a far-preferable solution.

      • mk3090
      • 7 years ago

      Personally I think the size is fine. I would prefer a case with some better cooling designed in rather than trying to fix the problem with a larger fan: [url<]http://www.logicsupply.com/products/nc200[/url<]

    • egon
    • 7 years ago

    Reminds me of laptop manufacturers shipping systems with modest fan speeds making for positive initial reviews/feedback, then slipping more aggressive, noisier cooling profiles into later BIOS revisions to deal with overheating complaints (result of dust, thermal interface material degrading, manufacturing variability, etc.).

    Not to suggest the circumstances were exactly the same with the NUC, the cooling being discovered as too marginal and fan behaviour changed before it hit the market.

      • willmore
      • 7 years ago

      Either I bought my laptop way late in this lifecycle or Fujitsu just did a crappy job of tuning the fan speed. The little Pentium B940 in here (i5 model in other version of the same chasis) turns on way too soon, runs way too fast, and doesn’t shut off until way too cool. Bump it up 20C and I bet the thing would never turn on unless I load the CPU.

    • moog
    • 7 years ago

    That hand. I’m getting all wrinkly just looking at it.

      • willmore
      • 7 years ago

      Kids these days. *shakes wrinkly fist*

      • UberGerbil
      • 7 years ago

      Those are Extreme GripLines™

      • Bensam123
      • 7 years ago

      I believe that’s a Scott wrinkly nub.

        • willmore
        • 7 years ago

        You must be farsighted. That’s clearly a hand.

          • Bensam123
          • 7 years ago

          I’ll call it a nub if I want to.

            • willmore
            • 7 years ago

            This is a family site, Sir!

    • Pettytheft
    • 7 years ago

    I’ll still wait for Rev.2

      • NeelyCam
      • 7 years ago

      Good idea. Haswell and USB3 is all that’s missing.

    • Bensam123
    • 7 years ago

    If a computer is designed around meeting a thermal throttling threshold normally something seems off. Maybe it’s a interesting way of dealing with thermal constraints, but it just seems like they’re trying to do too much with too small of a space.

      • just brew it!
      • 7 years ago

      Seems to me it depends on what you consider to be the “normal” use case. Throttling’s OK as long as it doesn’t significantly impact usability. Crashing is not OK (obviously).

      Perception is also somewhat dependent on how the behavior is described. Are we “throttling” when we hit thermal constraints, or are we going into “turbo” mode when those constraints are absent?

        • Bensam123
        • 7 years ago

        Yeah… obviously this turns relative pretty fast. Makes you wonder what then qualifies as a normal usage scenario… I’m sure AMD could squeeze a FX8350 into a similar size enclosure and then throttle it constantly, but that doesn’t necessarily mean you’re getting what you pay for.

        If I remember right they didn’t try playing any games on the NUC when they tested it… That’d be a pretty good indicator of a common usage scenario and it generates quite a bit of heat.

      • Logic Supply
      • 6 years ago

      Bensam123, we took a different tact, engineering a heatsink for the SSD in the enclosure we designed. (Logic Supply is an embedded systems designer) Initial tests at a 50% duty cycle suggests the Intel NUC continues to have thermal-related instability if you pair an Intel SSD and wireless. Our solution is a different cooling configuration and the heatsink. You can check out our test regimen on our blog.

    • tbone8ty
    • 7 years ago

    it basically comes down to intel poorly designed the nuc… there has been many of these types of mini pcs out for awhile now and ive never heard of these issues before.

    • NeelyCam
    • 7 years ago

    Why is this a problem only for the Thunderbolt version? Is the big bad TB chip producing too much heat even when it’s not being used (i.e., no power management)…?

      • Damage
      • 7 years ago

      Well, the other one doesn’t have a Wi-Fi card….

        • NeelyCam
        • 7 years ago

        No; they both have the same slot for a wireless module. I chose not to use mine since it has a Gb ethernet port, but I could have. That doesn’t explain the difference.

          • Damage
          • 7 years ago

          Hm. Well, the other one may have the same problem, but I haven’t confirmed it.

            • NeelyCam
            • 7 years ago

            From the article it sounded like you confirmed with Intel that only the TB version has the problem. I hope people don’t read it the same way and buy the GbE version, thinking it won’t have the same problem.

        • NeelyCam
        • 7 years ago
    • UberGerbil
    • 7 years ago

    What exactly was the nature of the 525’s temperature throttling mechanism, and its failure? Is it the controller that is overheating? I can’t say I’ve heard of a consumer SSD controller with this issue (or with any kind of thermal throttling feature) but maybe it’s just flown under my radar?

    Should SSDs be tested in high thermal environments as one aspect of their benchmarks for review purposes?

      • derFunkenstein
      • 7 years ago

      My guess is that the wifi controller was combining with whatever little heat the SSD controller creates to make for a hot spot. That’s most likely why the fan profile helped so much.

      • Damage
      • 7 years ago

      My understanding is that the SSD will slow itself down–presumably the controller, possibly also the NAND dice–if it’s getting too hot, in order to attempt to avoid a heat-related failure.

      In this case, I don’t think the heat was coming just from the SSD itself, but from the Wi-Fi card and the SSD together, and the chassis wasn’t expelling heat quickly enough to keep up.

      Then the pre-production SSD didn’t actually implement its throttling correctly, so we saw system locks instead of just storage slow-downs.

      This seems to me like a bit of an unusual event. One typically assumes the chassis will be designed properly for the thermal load–and SSD controllers are not high-wattage affairs with serious cooling needs.

      I’m not sure whether SSD controllers generally have logic to deal with worst-case thermal scenarios, but one would hope so. Could be interesting to investigate, but not in a major way since there aren’t many watts worth of heat involved.

        • just brew it!
        • 7 years ago

        [quote<]I'm not sure whether SSD controllers generally have logic to deal with worst-case thermal scenarios, but one would hope so. Could be interesting to investigate, but not in a major way since there aren't many watts worth of heat involved.[/quote<] Just guessing here, but normally your SSD isn't going to be operating practically on top of another really hot-running chip in a tightly confined space. So this may have been some new functionality they needed to add to the controller, which hadn't been thoroughly debugged yet.

          • Deanjo
          • 7 years ago

          With mSATA drives they are more likely to be tight confined space where heat could be an issue. I seem to recall a conversation with a Crucial engineer about a year ago where it was mentioned that their M4 mSATA drives would slow down if thermals started getting too high.

          • cygnus1
          • 7 years ago

          I would think any mSATA SSD should be designed with thermal management in mind as, I would think, large percentages of them are going to go into tight environments, like the NUC and inside laptops. This is assuming that the controller and flash can still operate accurately in the thermally throttled situation. I would rather the drive lock up and shut the system down than limp along and produce data errors.

        • willmore
        • 7 years ago

        Unless you were loading the SSD with something other than the wireless file transfer, how could it meaningfully effect a slowdown? Unless it slowed itself down below the rate that data was coming in over wireless, nothing would have changed with its performance.

        Do you have good copy speed measurements? For whatever value they can have by being over relatively unrepeatable Wi-Fi, that is.

        If it’s only being asked to write a few MB/s, what’s there to throttle? Maybe the controller (and the memory channels) weren’t being idled when not in use? if that were the case, then one would expect power consumption to improve with this update. Power consumption is a pretty important thing to get right in an mSATA device–which will probably be put in some mobile device.

Pin It on Pinterest

Share This