Apple makes 'last-minute decision' to use TSMC for 30% of 'A9' chip orders for next iPhone

2

Comments

  • Reply 21 of 49
    Not, necessarily true ... You can build something, say, to an A8X spec -- and get 30% yield (70% rejects).

    But, you can retest the rejects at, say, a lesser A8 spec -- and get an additional 50% yield ...

    So, the premier devices like the next iPhone have a low yield, but secondary devices like AppleTV, iPad Mini. etc. are much higher yields ...


    This is the way bleeding-edge semiconductor manufacturing works -- and those who can, plan for and take advantage of retesting at a lower spec.

    Normally, yes, as I noted a bit later.

    The problem is Apple doesn't have the variance that, say, Intel has. An A8X isn't a replacement for an A8, the die is too large. And Apple also doesn't have the clock speed variations either, the A7 was only 100MHz faster in the iPad Air than the iPhone and Mini.

    These chips are also cheaper to make.
  • Reply 22 of 49
    mrshowmrshow Posts: 164member

    These articles are hilarious in their absurdity. 

  • Reply 23 of 49
    I'd have to say it's pretty much impossible to mix 14nm and 16nm in the same product. Unless the design of the A9 is changed the chip would be physically a different size. So then you'd need to have to motherboard designs, one might end up now being slightly bigger. That might mean you now need a slightly smaller battery for the 16nm motherboard design because it no longer fits. You'll need 2 case designs as the mounting points on the board have changed, etc, etc. it just wouldn't happen.

    However having said that there's 2 iPhones currently. So it's not impossible that say the iPhone 6 Plus is on 16nm and the iPhone 6 is on 14nm. Or maybe there's a new 4 inch model so we end up with 3 models and that gets the lesser 16nm design. So with that in mind this could be possible. Or could simply be A9 is 14nm, A9X is 16nm. iPads have a lot more space and 16nm is still smaller than 20nm they are currently at.
  • Reply 24 of 49
    kkerstkkerst Posts: 330member
    30% for a first run is actually pretty good. Usually, yield rates are in the single digits for a new process - 14 nm in this case. It's a complicated recipe to get right before they make millions of them. Very common for this process to take many months to get the yield up.
  • Reply 25 of 49
    And it begins... In this instance, thanks Ming. Because of you a reference to a report flattering TSMC actually got published here.

    Now here is how this storyline will go until September... Hold on, I do not want to be censored again.

    Thanks, Cowen for your report last month!
  • Reply 26 of 49
    Quote:
    Originally Posted by AppleInsider View Post



    Facing poor yield rates from chipmaker GlobalFoundries, Apple has apparently made an eleventh-hour call as it solidifies its supply chain for the next-generation iPhone, opting to award nearly a third of "A9" chip orders to Taiwan Semiconductor Manufacturing Co.

     





    Well-connected analyst Ming-Chi Kuo of KGI Securities issued a note to investors on Wednesday, a copy of which was obtained by AppleInsider, revealing that Apple has apparently made what he called a "last-minute decision to recruit TSMC." Apple is said to have called an audible after partner GlobalFoundries continued to experience poor yield rates on production of the next-generation CPU.



    Specifically, GlobalFoundries' "A9" chip yield rate is said to currently be at about 30 percent yield rate, which is well below what Kuo said is a mass-production "basic requirement" of 50 percent.

    Apple is said to have turned to TSMC after partner GlobalFoundries showed exceptionally poor "A9" chip yield rates of around 30%.

    "Recruiting TSMC reduces supply uncertainties for Apple," the analyst said.



    Another factor in the decision, according to Kuo, are concerns from Apple that Samsung's chipmaking business may not be able to supply enough of its 14-nanometer design. That's because initial sales of the Galaxy S6 and S6 Edge have apparently been greater than expected, and may pull 14-nanometer orders away from Apple.



    Finally, Kuo also indicated that TSMC's competing 16-nanometer FinFET Turbo design has exceeded Apple's expectations in both yield rate and performance.



    Another industry analyst said much the same in a report last month. Citing a recent trip to Asia, Timothy Arcuri from Cowen and Company said strong yields and attractive pricing led him to believe that TSMC had secured a large portion of Apple's A9 order.



    Apple's 2015 iPhone update is widely expected to sport a next-generation processor based on a smaller and more efficient design. Multiple reports have indicated that Samsung will build the majority of "A9" processors for Apple's next-generation iPhone.



    Samsung had a stranglehold on Apple's mobile processor business, building all units for the iPhone until last year. That's when TSMC began contributing chips for the iPhone 6 and iPhone 6 Plus, using a 20-nanometer process for the A8 processor that powers Apple's flagship handsets.

  • Reply 27 of 49
    xiao-zhixiao-zhi Posts: 112member
    Quote:

    Originally Posted by tmay View Post

     

    Apple would have designs produced for both processes, and I believe that both TMSC and Samsung would have had to at least have a tapeout (preproduction) as part of the competition. I do think that Apple was always going to spilt the production anyway.




    So you think Apple would produce two versions of A9 for 16nm and 14nm FinFET processes and the dies would have equivalent performance?

     

    Of do you suppose TSMC is suddenly going to get a 14nm process working as they struggle to get 16nm op to yield?

     

    Frankly, it's more likely Samsung would help to get GF's problems solved if that is the issue. Apple is certainly splitting production of different designs but using dissimilar processes for a single SoC is a stretch.

     

    Keep in mind this is an unsubstantiated rumour, AppleInsider's breathless hype notwithstanding.

  • Reply 28 of 49
    loekfloekf Posts: 41member
    This Kuo guy should be fired and deported to Siberia.

    "can someone explain to me with a yield rate is? So if factory A is producing a 30% yield opposed to 50%, does that mean that it's producing only 30% of what apple needs? Or that 30% of the product is usable, and the rest is garbage? Thanks"

    During Semiconductor manufacturing the yield is the number of good dies ("chips") on a wafer. 50% yield is ridiculous. For logic the yield should be well above 90% otherwise there's seriously something wrong with either Apple's A9 design, the process or the modules itself.

    Usually, blocks like SRAM can be repaired using redundant cells so yield losses can be compensated for.

    I refuse to believe that Apple had basically two teams, one targetting a Samsung/GF 14nm process and a second backup team targetting a TSMC process. This would be very very costly and requiring lots of man power.
  • Reply 29 of 49
    wizard69wizard69 Posts: 13,377member
    xiao-zhi wrote: »
    That approximately correlates to the 25% allocated to Global Foundries in this leaked document published by semiwiki.
    Last I knew Global Foundries was not part of Samsung. They have been two independent companies.
    Or he could have taken that information and made an assumption it would be thrown to TSMC.

    Yet we can still wonder how TSMA will manufacture SoCs designed for Samsung's 14nm FinFET process using their 16nm FinFET process which they have struggled to get up to production yield and is 1 year behind schedule.
    In that regard I'm certain that each A series chip is designed against multiple production processes. For Apple it is all about covering your ass. You can't design for the bleeding edge without having hiccups from time to time. In other words 14 nam class tech isn't a sure thing from anybody anymore.
    So does AppleInsider have some special explanation of that?
    Talking out of their ass maybe? In the end though I'm certain that Apple has completed the design against multiple processes. Realistically they would have to to get the yield info required to allot manufacturing.
  • Reply 30 of 49
    wizard69wizard69 Posts: 13,377member
    can someone explain to me with a yield rate is? So if factory A is producing a 30% yield opposed to 50%, does that mean that it's producing only 30% of what apple needs? Or that 30% of the product is usable, and the rest is garbage? Thanks

    Think of it as a comparison of actual product starts to salable items coming off the production line. 30% could be considered terrible yields though I've worked in industries where such low yields where common.
  • Reply 31 of 49
    wizard69wizard69 Posts: 13,377member
    That's insane. What a huge waste.
    It isn't good but I've seen worst yields in the medical device industries.
    Is that because quality control measures renders them unusable?
    The process renders them unusable.
    Seems like such a huge waste of money to me if that much is being rendered unusable. I wonder if apple has to pay for the unused chips 

    In the end customers pay. A process with poor yields cost everybody more money. Yields are extremely important.
  • Reply 32 of 49
    tmaytmay Posts: 6,311member
    Quote:



    Originally Posted by Xiao-zhi View Post

     



    So you think Apple would produce two versions of A9 for 16nm and 14nm FinFET processes and the dies would have equivalent performance?

     

    Of do you suppose TSMC is suddenly going to get a 14nm process working as they struggle to get 16nm op to yield?

     

    Frankly, it's more likely Samsung would help to get GF's problems solved if that is the issue. Apple is certainly splitting production of different designs but using dissimilar processes for a single SoC is a stretch.

     

    Keep in mind this is an unsubstantiated rumour, AppleInsider's breathless hype notwithstanding.


    As others have pointed out, there are actually a number of products that the A9 is likely targeted for: two, maybe three iPhones, and;  two maybe three iPads. As long as the decision is made early, it wouldn't be much of a challenge. 

     

    While this is a rumor, it also demonstrates a future where Apple isn't necessarily going to get the same generation A series at the same process, or even the same foundry; capacity looks constrained.

  • Reply 33 of 49
    The following link is to a Barron's article that favors TSMC. The arti mentions Cowen who was mentioned by AI. The article is dated March 15, 2015.

    http://m.barrons.com/articles/BL-AFUNDSB-1949

    What adds humor to the pro-Samsung storyline is TSMC just announced a banner quarter. Here is a link to a Reuters article about this.

    http://finance.yahoo.com/news/taiwans-tsmc-books-65-pct-054243025.html?.tsrc=applewf

    The press is focused on TSMC's lower Q2 numbers due to Samsung dropping Qualcomm. Any good news about TSMC will get buried quickly to favor Samsung, which is supposed to now be on track to ship 32-35 million S6 phones between March - July and reach 50 million by year end. That 20 million preorder number is still being referenced as good news. ????
  • Reply 34 of 49
    xiao-zhixiao-zhi Posts: 112member
    Quote:
    Originally Posted by wizard69 View Post





    Last I knew Global Foundries was not part of Samsung. They have been two independent companies.

    In that regard I'm certain that each A series chip is designed against multiple production processes. For Apple it is all about covering your ass. You can't design for the bleeding edge without having hiccups from time to time. In other words 14 nam class tech isn't a sure thing from anybody anymore.

    Talking out of their ass maybe? In the end though I'm certain that Apple has completed the design against multiple processes. Realistically they would have to to get the yield info required to allot manufacturing.



    Samsung and Global Foundries have a strategic business relationship that includes maintenance of a common process, shared IP, shared process methodology and shared business.  Their recent bid for Apple business included proposals for business split between 3 foundries in Korea, Austin Texas and GF's New York plant.

     

    It is not that likely Apple would produce a version of the A9 in multiple process nodes because they have different performance characteristics.

     

    Instead what Apple has done is to split orders for different versions, e.g., A9/A9X between suppliers with different processes and to alternate overlapping generations between suppliers (how TSMC got the biz). The article I linked shows that exactly.

     

    Some people on this thread apparently unfamiliar with IC fabrication (or with a little bit of knowledge and big imaginations) you can slit orders between a 16nm and 14nm process and get the same results. Amusing idea, but not very realistic.

     

    AppleInsider is, can I say, partisan?

  • Reply 35 of 49
    Cut price deals = shopping around.

    It seemed obvious at the time, that with Samsung moving the S6 fully to Exynos, actually meant that Apple had signed for a much lower price, and thus a low priority deal.
  • Reply 36 of 49
    MacProMacPro Posts: 19,718member
    ecats wrote: »
    "Last Minute"... is ignorant of production planning and ramping. One doesn't just pick up a whole other chip foundry off the cuff.

    It's more likely that it was this way all along and the information, if true, was leaked only now.

    I agree and perhaps the two different versions are heading to two different Apple products, one being slightly thicker than the other.
  • Reply 37 of 49
    ecats wrote: »
    "Last Minute"... is ignorant of production planning and ramping. One doesn't just pick up a whole other chip foundry off the cuff.

    It's more likely that it was this way all along and the information, if true, was leaked only now.

    Ignorant of production planning? Why does Apple have to change fabs at the last minute?

    Most likely is Apple had both TSMC and Samsung/GF fabbing the A9 and have been monitoring yields for months now. After seeing the results they made a decision as to production spit. So Apple can make a last minute decision based on months of planning/production/testing.
  • Reply 38 of 49
    maestro64maestro64 Posts: 5,043member
    I do not believe this you can not have an SOC architecture on two completely different Fab lithograph. It will make the hardware design too complicated as well as the software to deal with it. This is why all companies award processor solutions to one supplier at a time.

    Notice where this is coming from, analysis who is trying to pump TSMC. if the industry believe TSMC is losing the business their stock will drop.
  • Reply 39 of 49
    karas11karas11 Posts: 32member
    Quote:

    Originally Posted by TheWhiteFalcon View Post

     

    I really doubt Apple would ship some A9's at 16nm and some at 14nm. So Ming-Chi is probably wrong. Again.




    Well, iPhone 6 has different storage. 16GB version used 100% MLC, other use mixed TLC & MLC.

    So Apple could do that again on iPhone 6s when it lunch.

    If 16GB version still base then it might use 16nm A9. 64GB might used mixed and 128GB version use 100% 14nm.

    This is problem for TSMC not for Apple. Apple does not care as long as they can get the quantity of A9 chip on time.

    If TSMC had 14nm then they might got 50%+ order from Apple.

     

    Some post mentioned about Sony PS3 cell processor, that does not the same case for A9.

    They use 7 cell because IBM could not deliver 8 enabled version. So Sony just changed to use 7 enabled version.

    Because they could not get the quantity they need.

     

    16nm & 14nm does not have any difference on chip design. Only 14nm could get little more performance due to smaller manufacturing.

    If A9 performance shows big difference due to 16 & 14nm then Apple might ordered 60%+ to Samsung.

    Since Apple decide mixed manufacturing processor version, we will find out soon enough when it arrived and run benchmark.

  • Reply 40 of 49
    Quote:

    Originally Posted by Maestro64 View Post



    I do not believe this you can not have an SOC architecture on two completely different Fab lithograph. It will make the hardware design too complicated as well as the software to deal with it. This is why all companies award processor solutions to one supplier at a time.



    Notice where this is coming from, analysis who is trying to pump TSMC. if the industry believe TSMC is losing the business their stock will drop.



    The same can be written about Samsung.

Sign In or Register to comment.