back to article Did you guess 2019 for Intel's 10nm chip ramp up? Congratulations

Meltdown? What meltdown? Intel's 50th birthday in July should be a hell of a bash as the chipmaker is flush with cash and reckons 2018 might well be a record-breaking year. Just don't mention two very little words: ten and nanometre. In a conference call with Wall St analysts on Thursday, to mark the release of Intel's Q1 …

  1. whitepines
    Devil

    Funny they mention "data centric", since they're quite interested in what you store on their systems. At least, considering they maintain *unremovable* vendor-level remote access capability via the management engine (note: this is different than the useful owner-level remote access capabilities), it seems they want to be a friendly big brother that can help make sure you only access and store what you are "supposed" to....

    Or was I supposed to draw a different conclusion from this?

    1. Will Godfrey Silver badge
      Unhappy

      Good points, but you made a mistake about the 'owner-level' bit. You are no longer the owner. They are, so everything's, fine, just fine.

      P.S. Do be careful what you say...

  2. Tom 64
    Pint

    Wow

    Intel have really screwed the pooch on 10nm. TSMC, GloFo, Samsung and more are already cranking out '7nm'. Breaking out some popcorn and looking forward to watching intel stock take a bath.

    1. Anonymous Coward
      Anonymous Coward

      Re: Wow

      Why would Intel's stock take a bath, or even be affected? They don't have any foundry business worth mentioning, so if they lost it to TSMC et al it wouldn't affect them. They will still sell just as many Intel Core <indecipherable number> CPUs because people who need a PC/server will buy one.

      Maybe AMD can get a brief leg up on Intel performance-wise, but they had that for a couple years in the Athlon vs P4 days, and it barely registered as a blip on Intel's stock.

    2. theblackhand

      Re: Wow

      Be careful comparing the processes at a "7nm is better than 10nm" level - TSMC/Samsung/GloFo 7nm processes are likely to have similar or worse characteristics to Intels 10nm in-terms of gate and interconnect pitch which will likely lead to similar performance.

      The big difference is historically Intel had a 3+ year lead over the rest of the industry on process technologies for a stable "next gen" process, which allowed them to roll out more conservative CPU designs than their rivals. That's where they've screwed the pooch more than once on 10nm already as that lead is now less than a year, allowing no real opportunities for further mistakes. .

      That Intel still can't get volume on 10nm after ~1.75 years, they may already be looking at 7nm chips although that will hurt them on volume until they retool the (likely economically unviable) 10nm fabs.

      1. Sykowasp

        Re: Wow

        Intel is 3 years late on 10nm now. It looks like it may be 4 years late if it isn't fixed by mid 2019 - it's clear from the CEO's comments that there is little confidence that they can fix the yield issues easily, even though he claims they understand it. This is very similar to how Intel gained that 3 year lead in the first place - when everyone else got stuck at 28nm for a couple more years than they would have hoped for.

        Their statements already clarified that they have to do 10nm before 7nm, because of sunk costs and their fab upgrade methodology. Going to 7nm would just introduce even more risk, if they cannot multi-pattern 10nm reliably still, as 7nm requires even more of that.

        Intel have historically used their fab advantage to push their CPU clocks higher than the competition (or lower power consumption), whilst having good [CPU] designs. This is evidently no longer the situation. In the future they may make up lost ground, but right now it looks like they will be behind in volume for a year (versus TSMC) and likely come in behind Samsung and GlobalFoundries as well. Also note that gate and interconnect width is not everything there is to a process - Intel use simple 1D routing that costs density even though the individual transistors are very nice, and there are other aspects as well.

        TSMC's 7nm looks very strong by the way, there is no guarantee that Intel's 10nm is going to equal it, never mind beat it.

  3. mark l 2 Silver badge

    "Breaking out some popcorn and looking forward to watching intel stock take a bath."

    I doubt that is going to happen anytime soon, if their share prices were going to take a bit hit it should have been back when the Meltdown vulnerabilities were revealed at the beginning of 2018.

    Yet they are laughing all the way to the bank by continuing to churn out chips with the flaws still present and fixed only with software that slows down performance and people are still buying them.

  4. Pascal Monett Silver badge

    Data-centric ?

    Oh, so that explains why you're not churning out the chips, then ? You're not a chip-maker anymore, you're a data center. Is there any area of computing that is not turning into a vast excuse to slurp MY data ?

    Oh well, there's AMD that's still a chip-maker. At least somebody is still doing what it is they're supposed to be doing.

    1. whitepines
      Facepalm

      Re: Data-centric ?

      Ehh, no. AMD has the same grubby "rented system" philosophy as Intel does. Look at the unremovable PSP (their counterpart to Intel's ME).

      I liked AMD back in the day, then they ran into some very hard times thanks to illegal anti-competitive practices from Intel (for which Intel was convicted, but not e.g. forced to give up patents gained as a result). They came back by being just as nasty as Intel with the whole "vendor gets access to your data, for your protection" stuff.

      Sad story, all around. The nastier you are to your customers, the more money you rake in....

  5. John Smith 19 Gold badge
    Unhappy

    So that's 90 atoms wide with a 9 atom thick oxide layer.

    The day of FET is drawing to a close.

    Tick tock.

    1. itzman

      Re: So that's 90 atoms wide with a 9 atom thick oxide layer.

      well the days of FET for chipping...

      Unless they go hybrid analogue digital... 16 voltage levels representing 4 bits in one prcessor cell.

    2. Anonymous Coward
      Anonymous Coward

      Re: So that's 90 atoms wide with a 9 atom thick oxide layer.

      And at least a few of them must be dopants to get N or P type material, and Murphy guarantees that at least some of them aren't going to be anything you want at all. Starting to feel cramped!

      1. Anonymous Custard Silver badge
        Headmaster

        Re: So that's 90 atoms wide with a 9 atom thick oxide layer.

        It's equivalent to a 9 atom thick oxide layer, but it's a high-K material, not SiO2.

        Hence physically it's rather thicker (and so less leaky), with the same electrical characteristics.

  6. Anonymous Coward
    Anonymous Coward

    He was at Tesla for just over 2 years. If one were a suspicious person one might think that is just long enough to avoid lawsuits for inside process/architecture knowledge being passed along from his AMD years.

    On another note, I hope Intel have fixed the architecture of their processors. Going to 10nm scale gives them the perfect opportunity, if they'll resolved the architectural issue.

  7. Chairman of the Bored Silver badge
    Pint

    Not bad for a firm was "stuck" fabbing DRAM

    In the dark ages before the microprocessor, Intel was an also-ran making dynamic ram chips. Federico Faggin's epic 4004 and later 8008 micros changed the world and transformed Intel. You cannot say that about many people, but he did.

    Sad note: F.F. (and you can see the F.F. on chip die) left Intel to launch the Zilog Z80 and Intel for years downplayed his role in creating the modern micro.

  8. Anonymous Coward
    Anonymous Coward

    is intel 10nm really shipping

    Krzanich kept saying 10nm is shipping. i dont think any products anywhere have 10nm in them. correct?

    i am sure they are shipping samples that cannot be used.

    intel: "our never to be functional 10nm is better than TSMC 10nm that has shipped 500M units!"

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

Biting the hand that feeds IT © 1998–2021