Samsung & TSMC begin volume production of Apple's next-gen 'A9' CPU for 'iPhone 6s' - report

2»

Comments

  • Reply 21 of 32
    wizard69wizard69 Posts: 13,377member
    melgross wrote: »
    Several reasons, I suspect.

    One is that until last year, Samsung has made every ARM chip for Apple, at least, from the A4 onwards, and possibly earlier. Apple has never had a real problem with Samsung. Two is that Samsung's 14nm process is supposedly better than that of TSMC's 16nm process. I've never felt comfortable with TSMC. Nvidia and ATI have had problems with them going back as far as I can remember. They're always late with a new process, and have problems for months afterwards.

    The NVidia and AMD problem are as much a problem with the GPU designers as it is with TSMC. The GPU manufactures tend to want as much performance as is possible often pushing what is advisable design wise. Speaking of late, compared to Intel or Global Foundries, TSMC is very reliable of late.

    As to which process is "better" I really don't think it matters too much to Apple. Everything I've seen indicates that Apple focuses on yields not ultimate performance. If the two processes can deliver to Apple SoC that operate with the same performance at acceptable yield rates I don't think you will see a preference one way or the other. The other factor of course is power usage and it isn't even clear if there is a huge difference here.
  • Reply 22 of 32
    cnocbui wrote: »
    Interesting to see in the news today that Intel can't do 10nm and are sticking with 14nm.  I imagine the 6S will sell very well but will represent the pinnacle for some time.

    If the 6s were to be a pinnacle for some time then Apple's iPhone growth would level off. You can trust that Apple has several generations of iPhones planned out beyond the 6s, and each one as compelling to upgrade to as always. As we near each new release the new standards become a bit clearer... who had ever heard of "force touch" a year ago?
    I suspect the only reason why Apple would bother with TMSC at 16nm is if Samsung don't have the capacity at 14nm for both the A9 and Exynos 7420\30.

    No there are many reasons for Apple to "bother" with TMSC:
    1. Apple uses TMSC to bid against Samsung to keep Sammy's profits thin.
    2. By buying from TMSC, Apple is helping a competitor to Samsung grow and afford R&D.
    3. Samsung is not Apple's friend, Apple is smart to develop and grow second-source manufacturing on critical components.
    4. and on and on...
  • Reply 23 of 32
    melgross wrote: »

    I believe Samsung has most of the production again this year. While Apple was luck with TSMC last year, I'd hate to see them push that luck. And while we don't know, a reason why Apple went to Samsung again this year could be related to TSMC having process problems again, as has been reported. I also would put too much credence on anything 10nm, as Intel is having major problems of its own, and if they're having problems, it's difficult to believe that others won't as well.

    IBM announced this week that they have successfully created a process for sub-10nm chips. While this new process is not ready for production, it has essentially halved the current 14nm production limit. Significantly, the research done by IBM was done in conjunction with a group of global partners (including Samsung).
  • Reply 24 of 32
    hentaiboyhentaiboy Posts: 1,252member

    Does it grate you fanbois that your iPhones beat with the heart of Samsung?

  • Reply 25 of 32
    hentaiboy wrote: »
    Does it grate you fanbois that your iPhones beat with the heart of Samsung?

    It depends on whether you think positively or negatively about the iPhone. Apple haters never speak about Samsung or LG (or other supplier) when criticizing the iPhone. Criticism is always goes to Apple, while credit is given to suppliers.
  • Reply 26 of 32
    Quote:

    Originally Posted by hentaiboy View Post

     

    Does it grate you fanbois that your iPhones beat with the heart of Samsung?




    Not as much as it grates me to see fucking trolls posting here due to some mental defect.

     

    Blocked.

  • Reply 27 of 32
    melgrossmelgross Posts: 33,510member
    tmay wrote: »
    I don't think "luck" was involved. Apple had both tapeout  and TMSC won, with capacity and yield in its favor. Samsung bet big on building out its fabs for 14nm and it is paying off for them. 

    I said lucky, because as I've already mentioned, TSMC has a history or not meeting its dates for new process, and not meeting its performance goals for as much as 6 months after the node comes out for them. So yes, Apple was lucky.

    Apple was also likely trying to tell Samsung that they don't have to rely upon them. A message that it's said Samsung heard. Now, Apple is mostly back with Samsung.

    Remember that app, of Apple's chips have been made in Samsung's Texas plant, which is a large one. But with Apple's sales expanding at a good rate, it's possible that that plant will have problems meeting all of Apple's needs over the next few years. Considering that Samsung went with their own SoC in all their top models this year for the first time, and Samsung's capacity in all of its chip plants are under pressure they've never been in before. We don't know if Samsung will reject Qualcomm's SoCs next year, if the problems are resolved, but if Samsung feels confident of their own chips, they may, and that will continue to pressure production.
  • Reply 28 of 32
    melgrossmelgross Posts: 33,510member
    smalm wrote: »
    It is not.
    Samsung's 14FF LPE had a small advantage over TSMC's 16FF.
    But 16FF was retired when 16FF+ went into production last month.

    To keep up with 16FF+ Samsung needs to replace 14FF LPE with 14FF LPP which they haven't done yet.

    Once Samsung switched to LPP both 16FF+ and 14FF LPP will have the same fin pitch, gate pitch, and metal pitch.
    We may see hardly any noteworthy advantage of one over the other.

    This is your opinion. The opinions in the industry are different. I'll go with theirs.
  • Reply 29 of 32
    gatorguygatorguy Posts: 24,176member
    melgross wrote: »
    Remember that app, of Apple's chips have been made in Samsung's Texas plant, which is a large one. But with Apple's sales expanding at a good rate, it's possible that that plant will have problems meeting all of Apple's needs over the next few years.
    Samsung has a partnership with Global Foundries too, who license the 14nm process from Samsung. That would allow them to make use of the GF facilities in New York State as well as Korea for overflow orders.
  • Reply 30 of 32
    melgrossmelgross Posts: 33,510member
    wizard69 wrote: »
    The NVidia and AMD problem are as much a problem with the GPU designers as it is with TSMC. The GPU manufactures tend to want as much performance as is possible often pushing what is advisable design wise. Speaking of late, compared to Intel or Global Foundries, TSMC is very reliable of late.

    As to which process is "better" I really don't think it matters too much to Apple. Everything I've seen indicates that Apple focuses on yields not ultimate performance. If the two processes can deliver to Apple SoC that operate with the same performance at acceptable yield rates I don't think you will see a preference one way or the other. The other factor of course is power usage and it isn't even clear if there is a huge difference here.

    Not really. Because TSMC couldn't deliver on newer nodes on time, both ATI and Nvidia had to use their newest designs, which did work just fine, on older, and slower nodes, which pushed power requirements up, and required lower clock speeds. Both of these were supposed to be resolved at the smaller nodes. Eventually, they solved the problems. But this happened several times.

    Global Foundries isn't my first pick for anything. But I would disagree about Intel. They are the most reliable producer, and have been now for a long time. Everyone acknowledges that they have the best fans.

    Yields are not the problem of the customer, but the fab. The fab and the customer arrive at a price and quantity, and it's then up to the fab to deliver. If they don't, then, just as in every other business, there are penalties. If the fab can't meet production requirements, then we see a customer using more than one fab, as backups. But you always want, as a manufacturer, to get all of your parts, as much as possible, from one manufacturer, though you usually want double, or for some items, triple sourcing in case of problems. In that case, where parts are custom made, you might be required to buy a certain minimum, or pay a fee instead.
  • Reply 31 of 32
    melgrossmelgross Posts: 33,510member
    IBM announced this week that they have successfully created a process for sub-10nm chips. While this new process is not ready for production, it has essentially halved the current 14nm production limit. Significantly, the research done by IBM was done in conjunction with a group of global partners (including Samsung).

    Yeah, as an R&D lab process, it worked on a small, not too complex, chip. But no one believes that this is more than proof of concept. Remember that Intel had a 10nm chip two years ago, and now, as I've been saying they would do, they've backed away from production for at least half a year. They're saying that 1@nm will co e in 2.5 years rather than 2 years. But they're confidently omitting that they've been behind for years, starting with 45nm. 14nm is almost an entire year behind their initial schedule. Some people think it's 6 months, but that's only because they revised release dates two times since they first announced it.

    So really, 10nm will be around 3.5 years after it was expected several years ago. If they see more delays, which is certainly possible, considering how long the new release date is from now, it could be 4 years.
  • Reply 32 of 32
    melgrossmelgross Posts: 33,510member
    gatorguy wrote: »
    Samsung has a partnership with Global Foundries too, who license the 14nm process from Samsung. That would allow them to make use of the GF facilities in New York State as well as Korea for overflow orders.

    I have little confidence in GF. If Samsung takes their plants over, that could change. Licensing a process doesn't mean that the manufacturing if it will go well. It's two entirely different things.

    GF has been struggling to find new customers. I don't see that changing any time soon.
Sign In or Register to comment.