Intel expresses 10-nm and 7-nm confidence in investor Q&A

Intel exec Murthy Renduchintala recently spoke at length about the company's future at the 39th Nasdaq Investor Conference, and the talk has been summed up in a transcript spotted by Anandtech. Renduchintala had much to say about the company's continuing plans for its 10-nm process technology and lifted the lid a bit on its plans for 7-nm production.

As TSMC has begun volume production of chips on its 7-nm FinFET process (roughly comparable to Intel's 10-nm process in its claims for density) and Samsung claims to have won the race to EUV insertion with its own similarly-dense 7LPP node, Intel finds itself in the unfamiliar position of being left behind in the race to the limits of silicon. Investors remain understandably curious about just how well the blue team is playing catch-up.

Murthy Renduchintala. Source: Intel

Renduchintala's comments on the company's work on 10-nm chips suggest interesting times lay ahead next year as the company begins introducing those products in volume. Renduchintala flatly contradicted the notion that the company has scaled back some of its ambitions for the 10-nm process in order to boost yields. In response to one question, the exec noted that "the power and performance and transistor density targets that we set in 2014 remain the same." That could mean Intel's first 10-nm products will still achieve a roughly 2.7x density improvement over its own 14-nm process.

Renduchintala also shed some light on the company's plans for its 7-nm process technology as part of the Q&A. The company's past comments about 7-nm suggested that the fate of that process partially rested in lessons that it needed to learn from ramping 10-nm. Indeed, Renduchintala said that even though Intel's 7-nm development lies in "a separate team and a largely separate effort," that presumed team has taken "a lot of lessons out of the 10-nanometer experience as we defined that and defined a different optimization point between transistor density, power and performance and schedule predictability."

Intel's Chandler, AZ facilities, where its 7-nm Fab 42 will be located

One takeaway from those lessons appears to be a somewhat-less-ambitious scaling goal for 7 nm. Renduchintala notes that Intel 7-nm parts will return to something closer to a traditional 2x scaling goal—about 2.4x, if past comments by Intel execs hold. While that isn't news on its own, Renduchintala's statements confirm that the company's ambitions for 7 nm might wisely be more conservative in light of its 10-nm woes. Renduchintala also reiterated the fact that 7 nm will be Intel's first process with EUV lithography.

As far as when we'll see 7-nm chips, however, Renduchintala only said that the company is "very, very focused on getting 7 out according to our original internal plans"—a timeline that has never been publicly disclosed, to my knowledge. Advanced manufacturing nodes of the future are expected to have long lifespans to allow companies to recoup their investments in getting their fabs up and running, however, so it seems likely that 7-nm silicon is a long while off if Intel remains committed to ramping and refining 10-nm parts and bringing them to market across all of its product lines.

Comments closed
    • jarder
    • 10 months ago

    Meh, Fool me once, shame on you; fool me twice etc.

    • BorgOvermind
    • 10 months ago

    In plain English:

    10nm epically failed and of course we learned a lot of lessons out of the 10-nanometer experience, as in how to actually continue to innovate instead of halting it because we are market monopoly.

    Therefore, we promise the 7nm will be better than the ‘on-paper better’ we usually promise.

    • Chrispy_
    • 10 months ago

    If a company [i<]doesn't[/i<] say that everything is going to be great at their [b<]investor Q&A[/b<] then they're publicly committing seppuku.

      • tacitust
      • 10 months ago

      Except a company telling investors something that is materially untrue is also known as securities fraud.

        • NoOne ButMe
        • 10 months ago

        yes, but they’ll spend tens/hundreds of thousands of dollars of time with lawyers/etc. to make sure they have it spun optimally where they reveal everything, while telling just as much/little as required.

        A lot of it could be fixed by requiring answers to questions to actually answer them, but that has plenty of problems as a concept.

        • Srsly_Bro
        • 10 months ago

        Didn’t Intel tell investors that it was all good regarding 10nm progress?

        • Chrispy_
        • 10 months ago

        Which is why weasel-wording is the order of the day with any and all investor discussions.

    • Wirko
    • 10 months ago

    [quote<]Intel expresses 10-nm and 7-nm confidence[/quote<] Confidence expressed in units of length. Nice touch, Jeff.

      • K-L-Waster
      • 10 months ago

      All Ice Lake CPUs will require 128GB of on-package HBM3?

      • Srsly_Bro
      • 10 months ago

      I lol’d.

    • turtlepwr281
    • 10 months ago

    I’m just holding out for 10nm+++++

    • Srsly_Bro
    • 10 months ago

    Raja will soon execute Dr. Lisa Su’s master plan.

    • NTMBK
    • 10 months ago

    It’s all fine! Nothing to see here!

    • blastdoor
    • 10 months ago

    It sounds like the basic story remains the same — 10nm HVM sometime in 2019; 7nm later, but not too much later.

    I’ll guess 7nm HVM in 2021. Other guesses?

      • DancinJack
      • 10 months ago

      Maybe even “holiday” 2020 for 7nm.

        • Antimatter
        • 10 months ago

        Then why not cancel 10nm altogether if it’s only in use for 1 year?

          • blastdoor
          • 10 months ago

          Perhaps it will continue to be used in parallel with 7nm?

          Maybe chiplets on 7nm, IO chip on 10nm

          • Sahrin
          • 10 months ago

          Everything in the product stack moves forward (‘downward’) at the same cadence.

      • DavidC1
      • 10 months ago

      2021 HVM seems very unlikely for 7nm. Maybe 2022. Add a year for products.

      They are planning Comet Lake, which is yet another Skylake for -S, -H, and -U parts. That means at best we might see -Y Icelake chips towards holiday 2019, and real launch for Icelake is 2020.

      2020-1st Gen 10nm
      2021-2nd Gen 10nm
      2022-7nm? I don’t think so. They struggled considerably on 10nm. I think we’ll see a 3rd Gen 10nm
      2023-7nm? Maybe.

      Remember the presentation(by none other than Intel) that showed how number of companies fabricating chips decreased exponentially in the last few years? They had TSMC, GF, Intel, and Samsung. GF is out of the race. By the way they are going Intel might be next.

        • DancinJack
        • 10 months ago

        lol OK David.

          • DavidC1
          • 10 months ago

          I’m glad I entertained you, but its more truth than its not.

          They’ve been unable to deliver 10nm for over 3 years now. There’s no reason to expect 7nm can come faster.

            • NoOne ButMe
            • 10 months ago

            10nm issues are mostly confined to that process.
            Like how TSMC’s 40nm issues were just 40nm.
            Same for TSMC 32nm issues just being 32nm.

            • DancinJack
            • 10 months ago

            Nah bro, 10nm mistakes = INTEL IS DOOMED FOREVER.

            • Srsly_Bro
            • 10 months ago

            And 20nm…

            • NoOne ButMe
            • 10 months ago

            The 20nm process was fine for what is was.

            Just what is was wasn’t attractive for most chips, as power density went way up.

            It wasn’t broken enough to be canceled like TSMC 32nm. And like what Intel should/will do for 10nm.

            • blastdoor
            • 10 months ago

            Interesting comparison — will Intel end up fabbing more or fewer chips on 10nm than TSMC fabbed on 20 nm? Note that there were about 220 million iPhones 6 and 6+ sold, and that’s not the only product that used 20nm chips from TSMC (but probably the highest volume product by far)

            • NoOne ButMe
            • 10 months ago

            Are we counting chips actually shipped to consumers/corporate market, or wafers?

            TSMC will surely win on devices shipped imo—If Intel’s 7nm estimates are correct.
            On Wafers run through the fabs, Intel might have already outdone TSMC.

            But yeah, I think the best comparison is TSMC 32nm with Intel’s 10nm. The difference being TSMC’s leadership knew to take the short term negative hit to stop wasting money/time on it. While Intel’s leadship probably will end up spending 5-10 billion dollars and years trying to fix 10nm with nothing to show.

            • Stonebender
            • 10 months ago

            Sure there is. The whole problem with 10nm is the insane amount of multi-patterning required. Since the EUV lasers are a whole magnitude smaller, this will be less of an issue.

        • NoOne ButMe
        • 10 months ago

        a year seems high. you don’t need HVM to validate a product. Which is where much of the delay comes from.

        We’ll see products about 6-7 months after HVM start, and the OEM products at the next big refresh, iirc the refreshes happen about twice a year.

    • Redocbew
    • 10 months ago

    When asked questions on a topic like this I’d like just once to see an exec say “No way man, we’re totally hosed.”

      • K-L-Waster
      • 10 months ago

      When asked about the company’s prospects, the CEO and CFO in unison shouted “sell all your shares NOW!!”

        • Srsly_Bro
        • 10 months ago

        And remember the former CEO sold quite a few shares before reporting that AMD could regain significant market share. But don’t be alarmed, it was a scheduled sale.

          • Spunjji
          • 10 months ago

          Indeed! Scheduled to happen just before the also-scheduled (and much delayed) announcement of the Spectre and Meltdown vulnerabilities. Scheduling is great!

            • Srsly_Bro
            • 10 months ago

            The only thing Intel can meet a schedule on: delaying disclosure of material information to coincide with CEO stock sales and misleading investors.

      • Chrispy_
      • 10 months ago

      My inner voice read that in the voice of Ted Buckland from Scrubs.

    • Mr Bill
    • 10 months ago

    Intel’s [url=https://www.theverge.com/2015/8/12/9136311/raymond-chandler-google-letters<]Chandler[/url<] Facility... [quote<]Did you ever read what they call Science Fiction? It's a scream. It is written like this: "I checked out with K19 on Adabaran III, and stepped out through the crummaliote hatch on my 22 Model Sirus Hardtop. I cocked the timejector in secondary and waded through the bright blue manda grass. My breath froze into pink pretzels. I flicked on the heat bars and the Bryllis ran swiftly on five legs using their other two to send out crylon vibrations. The pressure was almost unbearable, but I caught the range on my wrist computer through the transparent cysicites. I pressed the trigger. The thin violet glow was ice-cold against the rust-colored mountains. The Bryllis shrank to half an inch long and I worked fast stepping on them with the poltex. But it wasn't enough. The sudden brightness swung me around and the Fourth Moon had already risen. I had exactly four seconds to hot up the disintegrator and Google had told me it wasn't enough. He was right."[/quote<]

    • chuckula
    • 10 months ago

    Intel is announcing its bankruptcy next Tuesday…… [b<]NANO-CONFIRMED[/b<]

      • albundy
      • 10 months ago

      meh, companies come and go. easily replaced. i will send flowers.

      • Unknown-Error
      • 10 months ago

      Master Chucky!

      You gotta do better than that. I fear you are slipping… 🙁

Pin It on Pinterest

Share This