Intel denies report claiming that it “killed off” its 10-nm process

A bombshell report from Charlie Demerjian at SemiAccurate this morning claims that Intel has stopped work on its long-delayed and long-beleaguered 10-nm production process. SemiAccurate is presently down, but the report was apparently troublesome enough that Intel felt the need to weigh in on Demerjian's claims by way of its official Twitter account.

For its part, the company says that 10-nm is alive (if not yet well). The company's tweet asserts that "[w]e are making good progress on 10nm. Yields are improving consistent with the timeline we shared during our last earnings report." If that's the case, the company would begin shipping 10-nm silicon in the holiday season of 2019.

As SemiAccurate only releases full articles to subscribers, we can't assess the breadth or depth of Demerjian's sourcing. Intel would be in seriously hot water with regulators if it stated an outright falsehood that could materially affect the value of its stock, however, so we doubt the company is blowing smoke in response to this report.

Past statements by Intel have suggested that the work involved in getting 10-nm silicon working is necessary for its foundries to make the subsequent leap to 7-nm fabrication, as well, so it seems unlikely that the company is simply abandoning its 10-nm plans entirely. Presuming Intel is telling the truth, we'll just have to wait and see what form 10-nm silicon takes if or when it finally comes to market.

Comments closed
    • DPete27
    • 1 year ago

    By the time they get 10nm ready, everyone else will be on 7nm, so they might as well just skip 10 and go straight to 7.

    • tipoo
    • 1 year ago

    It may not be cancelled, but it will also be different from the 10nm process we were initially shown, with compromises made to density to get it to viable yields. Wonder how it will compare to TSMC 7nm when all is said and done.

    10nm in its first incarnation was way more ambitious than 7nm if you looked at multi-patterning and FEOL vs BEOL and everything, but I wonder if they’ll de-tune it to the point of practically only matching 7nm to get it out, even after 7nm is shipping en masse.

    • Klimax
    • 1 year ago

    So massively unsurprised that “SemiAccurate” might be terminally Inaccurate in its reporting on Intel…

    I’d wait on more completely independent confirmations before trusting Mr. Demerjian on anything Intel-related.

    • Unknown-Error
    • 1 year ago

    Charlie is a bit of hit and a miss. I’ll remain skeptical.

    • ronch
    • 1 year ago

    It’s a good thing Intel has a very good CPU architecture that still runs great on an aging (Intel) process and that they have tons of money to keep them afloat. Delays of this kind would’ve finished other companies.

      • DancinJack
      • 1 year ago

      Also, they had a multi-year advantage when it comes to process tech. Their 14nm++ is still pretty amazing.

    • HERETIC
    • 1 year ago

    Intel 10nm is borked- there’s no doubt there.

    It’s all to do with who’s doing the interpretation.
    10nm is so badly broken, they are re-doing part at 12nm.
    So to say 10nm is dead, would be CORRECT…………..
    If Intel calls its 12nm 10nm, and we have working silicone
    then 10nm is WORKING FINE………………

      • blastdoor
      • 1 year ago

      Actually….. the process that the foundries call “10 nm” is more similar to Intel’s 14nm.

      So Intel could perhaps just rejigger 14nm one more time and call it 10nm.

    • watzupken
    • 1 year ago

    Most likely killed off by quickly transitioning out of it. It is a good idea they cut loss instead of throwing more money at the problem only to get a poor ROI in the end. Whatever Intel is doing with their 10nm, it is clearly not working or at least not worth pursuing considering its been delayed for no less than 3 years now.

    • TheRazorsEdge
    • 1 year ago

    Since their executives have told investors that 10nm is going to happen, they pretty much have to use it or face an SEC investigation that will expose every single problem they’ve had with it. And the SEC really doesn’t like it when company officers lie to investors.

    Intel 10nm will evaporate faster than a night with Bill Cosby. The absence of trouble reports on their 7nm process suggests it may be ready within a year of 10nm.

      • DancinJack
      • 1 year ago

      “Absence of evidence is not evidence of absence!”

        • TheRazorsEdge
        • 1 year ago

        Perhaps, but at the same time it is very hard to believe that Intel has leaks left and right about one process with absolutely no leaks on another.

        I strongly suspect 7nm is in much better shape and will be much closer to its roadmap.

          • DancinJack
          • 1 year ago

          Yeah I don’t completely disagree that 7nm is probably in better shape than 10nm was at the same point in its lifespan, but until I see products that actually leverage advantages of either I just don’t really care much.

      • Ninjitsu
      • 1 year ago

      “Intel 10nm will evaporate faster than a night with Bill Cosby.”
      what the shit ?

    • blastdoor
    • 1 year ago

    When I saw this headline my first thought was “uh oh, Bloomberg is at it again!”

    • DancinJack
    • 1 year ago

    Intel has an earnings report on thursday. Should be fun.

      • albundy
      • 1 year ago

      i doubt it would include this new info.

    • Freon
    • 1 year ago

    [url<]https://i.imgur.com/BXrVA3k.png[/url<]

    • DeadOfKnight
    • 1 year ago

    I’m skeptical that 10nm and beyond will even bring the level of improvements we’re used to seeing. Low power/efficiency still has some room for improvement, but we’re probably going to need bigger chips to get double digit improvements for mainstream products. Market segments have already been climbing in price. My guess is that maybe one last big leap is in store for us if they get EUV ironed out and they finally start using 450mm wafers. Moore’s law will be dead soon, despite all Intel’s efforts to convince us otherwise. It’s no wonder Intel is doubling down on parallel processing.

      • NoOne ButMe
      • 1 year ago

      Only Intel and Apple could benefit from 450mm wafers right now (factoring in R&D cost to get 450mm equipment). In my estimation.

      Extra one time costs means if youโ€™re not at a certain volume 300mm is cheaper.

      And that excludes the R&D cost.

      AMD might be big enough to benefit soon, but… R&D costs

      • blastdoor
      • 1 year ago

      Moore’s Law is dead, no “soon” about it.

      But that doesn’t mean progress ends, just that it happens more slowly.

      And I agree — fewer and fewer products will be taken to new nodes, at least initially, because it just won’t be worth it. AMD on TSMC 7nm is an early example — GloFo couldn’t afford to move to 7nm just for AMD — AMD is riding off of investments made to serve Apple. I suspect the lag between Apple and AMD will increase on the adoption of new nodes.

        • DeadOfKnight
        • 1 year ago

        By definition, yes, it’s dead. But the slow march of progress in terms of process tech is about to hit the top of the plateau in the next few years. The ball is about to be entirely in the court of the architectural engineers to bring greater perf/$.

          • Magic Hate Ball
          • 1 year ago

          And efficient programming will reward those who can make their software run better on the same hardware when we are no longer able to depend on hardware speed increases.

    • freebird
    • 1 year ago

    Actually, they stopped 4 years ago and just forgot to tell us… ๐Ÿ˜›

    • Wirko
    • 1 year ago

    It looks like a parody, swims like a parody, and quacks like a parody. A good one at that.

    • Shinare
    • 1 year ago

    Bring on the 500 core x86 processor!!!

    • Krogoth
    • 1 year ago

    Unless Intel manages to undergo drastic changes with its 10nm process. It is going to be only good for small and relatively simple ICs.

    Intel tried too many things at once with its 10nm process (Cobalt and new layering techniques). It is a nothing short of a miracle that they got working product at all.

      • maxxcool
      • 1 year ago

      Agreed, it almost feels like a really spendy experiment… and or “WTH let’s try it” knee jerk reaction to AMD.

        • psuedonymous
        • 1 year ago

        More like “we have so little competition, we may as well spend more on R&D than on moving processes to production”.

    • maxxcool
    • 1 year ago

    Hmm. if you make chipsets on 10nm, is 10nm still considered ‘canceled’?

    • Neutronbeam
    • 1 year ago

    Jeff…You know I can’t believe any TR rumor story unless you add the qualifier term “purported” a few times. To think that rumormongering has degraded to the point that you can’t even get a single “purported”–these are dark times indeed. ;->

      • K-L-Waster
      • 1 year ago

      Anything from Semi-Accurate comes automatically with a silent “purportedly.”

        • psuedonymous
        • 1 year ago

        And appended by “confirmed through exhaustive tea-leaf analysis and goat entrail examination at the full moon”.

      • derFunkenstein
      • 1 year ago

      “Purport” is better used to describe claims made by manufacturers, and that’s why it appears mostly in “so and so releases the new something” type of article.

      If anything is “purported” in this story, it’s Intel’s side. Intel purports that it’s continuing work on its 10-nm process.

        • derFunkenstein
        • 1 year ago

        somebody doesn’t like that words should be used correctly. ๐Ÿ˜†

    • NoOne ButMe
    • 1 year ago

    This sounds untrue. Or at least a deceptive title (Intel kills 10nm… by calling it 7nm like foundry?)

    Although given Charlie’s record on Intel’s 10nm, it’s hard to ignore.

      • ptsant
      • 1 year ago

      The title is probably clickbait, but in essence it’s not unlikely that the effort has shifted from 10nm to 7nm.

        • NoOne ButMe
        • 1 year ago

        Clickbait for what?
        The zero ads they run?

          • DancinJack
          • 1 year ago

          Still gotta get visitors to get subscribers. Though obviously it’s not the usual model of SERVE ALL THE ADS as most sites use for revenue.

            • NoOne ButMe
            • 1 year ago

            which is why they put the news above the paywall?

            Like, if this is clickbait it defeats it’s own purpose.

            • ptsant
            • 1 year ago

            Of course not. If you read the linked article you will see that the juicy part is for investor/business subscribers paying something like $1000 a year.

            • NoOne ButMe
            • 1 year ago

            The last paragraph basically tells you everything?

            I think there has been some Clickbaity stories on 10nm published on Semiaccurate. I don’t think this is one of them, well. Not if you read what is there.
            About companies dying (paraphrase) due to Intel’s 10nm failures, without stating companies besides they’re large… That’s subscriber-baity as hell.

            final paragraph (above the fold): “The knifing of 10nm shows that Intel is finally willing to do the right things for the right reasons even if it costs them some short term pain, it is the first adult decision we have seen from the company in several years. Let us walk through the reasons why it is a good thing…”

      • Beahmont
      • 1 year ago

      Charlie has been predicting the ‘imminent’ demise of Intel and Nvidia for literally the last decade.

      He doesn’t get points for finally being in the ballpark when Intel actually has real trouble.

        • NoOne ButMe
        • 1 year ago

        You missed the part where he has pretty much been breaking all the 10nm news on Intel (and been accurate).

        Charlie’s predictions (where the problems can seep in) tend to be his analysis of the data. Rather than the data he gathers.

        • Zizy
        • 1 year ago

        His problem in those cases was the same as of analysists: he had some good data and made wrong predictions based on interpretation of that data.

        For NV: He believed that because APUs are finally coming and the released ones even good enough for lowend gaming, AMD and Intel will push towards powerful gaming APUs and kill dGPU market. However, such APUs only came for consoles. Imagine seeing PS4’s chip as the desktop part, or even the X1’s (minus esram). It would destroy midrange dGPUs because all the prebuilts would be coming with that instead – about the same performance and vastly cheaper and simpler for OEMs. He of course also failed to consider dGPUs would be used for other purposes – nobody expected crypto craze and the AI stuff taking off the way it did.

        As for Intel’s demise: I don’t think he predicted that, but yeah, he, like everyone else, expected much more from the ARM servers. He has been telling for a while that 22 nm isn’t rosy, 14 nm had more troubles and that 10 nm has been completely broken. He also said Optane is broken and that their low-power ambitions are dead and so forth… But most (or maybe even all) of these “Intel’s demise” stories have turned out true in the end.
        Only the most recent stories – since Zen launched – have stuff like “Intel is in deep shit”. We shall see if this ends up true in a few years, but right now I find it reasonably likely based on 10 nm issues – AMD will have 7 nm TSMC vs Intel’s 14 nm – TSMC should be much denser and lower power.

          • K-L-Waster
          • 1 year ago

          With NV he also posted a number of articles claiming their financials were faked and they were on the verge of collapse, and that Pascal was vapourware (that last less than a month before the 1080 and 1070 shipped).

          “Semi” is somewhat overstated.

    • Shobai
    • 1 year ago

    [quote<]SemiAccurate is presently down[/quote<] Really, chuckula? This time you've gone too far! [/wink]

      • chuckula
      • 1 year ago

      I like to think of it as performing a public service, so I don’t even charge for it!

        • BurntMyBacon
        • 1 year ago

        Since you don’t charge, will you accept up thumbs in lieu of fiscal compensation?

    • cygnus1
    • 1 year ago

    Even if what they say is true, I think 10nm is going to be short lived in actual production, as I would imagine Intel has got to be prioritizing 7nm behind the scenes at this point if they’re at all interesting on keeping their manufacturing process edge over competitors.

      • DancinJack
      • 1 year ago

      I was going to post something similar. I don’t really know how long 10nm is going to be in volume production so I won’t guess there, but I’d have to imagine as soon as Intel knew they were in real trouble with 10nm, they started with tweaks and/or 7nm.

      IIRC they had some slides/info the past couple years that said first gen 10nm wouldn’t even perform as well as 14/+/++ (i can’t remember which one specifically), but density would be increased.

        • blastdoor
        • 1 year ago

        Am I imagining things or did Intel say that people are mistaken if they think delays in 10nm will translate into delays in 7nm?

        When I heard (or imagined I heard) that, I inferred that 10nm won’t be with us long or used for much…. which could be expressed as it being “killed off,” though I can understand that Intel’s pride would prefer not to look at it that way.

          • DancinJack
          • 1 year ago

          It’s changed many times along the way I think. At one point they were talking 10, 10+, and 10++. The timeline is so screwed though. They original said ~2016 for 10nm. It’s gonna be like, holiday 2019 before there are a reasonable number of 10nm CPUs in the wild.

          [url<]https://hothardware.com/news/intel-details-advanced-10nm-node[/url<] They also said at one time Ice Lake would be 10+. [url<]https://www.anandtech.com/show/11722/intel-reveals-ice-lake-core-architecture-10nm-plus[/url<] I'm not sure what the plan is anymore tbh. I don't think they ever said explicitly that 10nm would be a fast (relatively) transition node to 7nm, but it might end up that way and that's fine.

      • ptsant
      • 1 year ago

      I also think that they are not outright lying. Intel is already claiming 10nm output with the highly not successful i3-8121U, the sole chip to be manufactured in that process, probably so that they can claim some compliance with their roadmaps.

      In that sense, the 10nm process wouldn’t have to completely stop but if all major server chips and desktop chips are moved to 7nm directly that would effectively mean a massive shift in priorities. This is not such a far-fetched scenario but it is a risky bet: other they gain time by skipping 10nm or they lose a lot by staying to long at 14+++ while others are at 7nm.

        • cygnus1
        • 1 year ago

        Exactly. I don’t think they’re outright lying either, just covering their rears on what they’ve said publicly about 10nm for a long time. I think they’re probably pushing 10nm and 7nm equally hard at this point because 10nm should’ve already been behind them by now. And so if they can get 7nm out reasonably quickly after 10nm, it would only be prudent to not upgrade a ton of fab capacity to 10nm and instead target the majority of it to go from the 14nm varieties to 7nm.

      • tsk
      • 1 year ago

      [quote<]they're at all interesting on [s<]keeping[/s<] regaining their manufacturing process edge over competitors. [/quote<] FTFY

        • cygnus1
        • 1 year ago

        yup, good point.

        #truth

Pin It on Pinterest

Share This