Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/taiwan-news-media-claimed-apple-has-decided-to-use-tsmc-to-produce-at-least-80-of-a9.5157/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Taiwan news media claimed Apple has decided to use TSMC to produce at least 80% of A9

hist78

Well-known member
In the past several days I have read some interesting news articles from Taiwan and most of them are written in Chinese. Some of them have presented detail information. I must admit these reports might be based on rumors and I don't have a way to verify them. So please don't harsh on me or even ask me to prove them with some insider's top secret (I can't because I don't have).

?? ???????? | ?? | ?????
CTIMES - MIC??????A9????:???,??,A9,MIC,???,TSMC,??,Samsung,??,???,??,???
MIC?????????A9?? | ???? | ??????? CNA NEWS
TSMC to secure most orders for Apple's A9 processor: MIC | Economics | FOCUS TAIWAN - CNA ENGLISH NEWS


Some highlights from those articles:

1. TSMC has won about 80% of Apple A9 orders and Samsung will get the remaining 20% orders if Samsung can meet certain Apple's yield rate expectation.

2. Apples decided to use TSMC 16nm FinFET+ for 80% of A9 production is due to TSMC's better yield and better quality. Apple determined that it's just too much risk and too much uncertainty to use Samsung's 14nm manufacturing process.

3. TSMC's Fab 14 P7 plant in southern Taiwan will produce Apple's A9 processors with 16nm FinFET+ process.

4. TSMC will have about 50,000/month wafers capacity for 16nm FinFET+ ready by the end of June 2015. Among them. 17,000 wafers capacity is from reassigning existing 20nm equipment and about 33,000 wafers will be produced by new acquired equipment.

5. TSMC is in full speed to purchase and install new equipment in order to meet the tight 2015 production schedule.

6. After securing Apple's A9 16nm FinFET+ order, TSMC has started shifting focus to 10nm process.

7. According to the number provided by those articles, next year TSMC will spend $11.5 billion in capital expenditure. It will be $11.3 billion for Samsung and $11 billion for Intel in 2015. This will be the first time TSMC spends more in CAPEX than Samsung and Intel.


<script type="text/javascript" src="//platform.linkedin.com/in.js">
lang: en_US
</script>
inShare1​
<script type="IN/Share+init" data-counter="right"></script>
<script type="text/javascript" src="//platform.linkedin.com/in.js">
lang: en_US
</script>
 
Last edited:

From the above news story of Dec 26:

Samsung Electronics Co. of South Korea could share the Apple orders with TSMC because the U.S. consumer electronics giant tends to lower production risks by not relying on a sole supplier, the MIC said.

But Apple's orders to Samsung will only be a small part of the total and will be determined by the yield rate of the South Korean firm's 14nm FinFET processor


The conclusion is similar to what said over a month earlier on Nov 19, 2014:

Apple’s A9 processors will be mostly built at TSMC. Samsung may, just may, get a small portion of the orders, mainly because Apple may wish to have a second source as leverage for price negotiation with TSMC.

https://www.semiwiki.com/forum/f293/samsung-strikes-chip-deal-apple-4864.html#post16870
 
Thank you both very much for the research.

I would like to contribute four things for discussion:

First, contrary to what the media is saying Apple has never in the history of their SoC design second sourced. Up until the A8 Samsung had manufactured all Apple SoCs as a single source. The A8 and A8x are manufactured by TSMC as a single source. Of course there was pricing negotiations and Apple did get a very competitive wafer agreement from TSMC by pitting them against Samsung at 20nm. Remember, at the time the contract was negotiated it was not known that Samsung would not yield at 20nm. Since Samsung did not yield at 28nm that was certainly a sound prediction however.

Second
, it was my understanding that two wafer agreements were negotiated in 2014: Samsung would manufacture the A9 and TSMC would manufacture the A9x. This was mostly due to Samsung being ahead with the 14nm PDKs by 3-6 months but also that the Samsung 14nm LP was in fact lower power (based on simulations) than the first version of TSMC 16nm. This certainly is considered second sourcing but to be clear these are two separate designs. Remember, the A8 is a low power design and the A8x is high performance with a much higher transistor count.

Third, while wafer agreements are negotiated during the design phase there are manufacturing milestones that must be met to mitigate the risk of using leading edge process technologies. Considering the fact that Samsung did not yield properly at 28nm and Samsung also had problems with 20nm, you can bet that the 14nm manufacturing milestones were strongly worded and monitored.

Fourth
, for Apple to include 14nm silicon in the iPhone7 (September of 2015) they will need production wafers in Q2 of 2015 (TSMC started shipping production 20nm wafers to Apple in June of 2014). In September of 2014 I heard multiple reports of Samsung 14nm yield problems which were somewhat confirmed in October of 2014 by an equipment manufacturer who spoke publicly about it during a quarterly conference call.

Bottom line: I certainly believe that Samsung told the truth about winning the Apple A9 SoC business when I first read about it. I also believed the rumor that TSMC won the A9x with the 16nm FF+ process. I'm not privy to the wafer agreements of course but the consensus amongst the fabless semiconductor ecosystem was that this was in fact the case. I will do another poll amongst the fabless folks after the holidays to get the final consensus on who will manufacture what for the next Apple iProducts.

<script src="//platform.linkedin.com/in.js" type="text/javascript">
lang: en_US
</script><script src="//platform.linkedin.com/in.js" type="text/javascript">
lang: en_US
</script>
<script type="IN/Share" data-counter="right"></script>
 
Thank you both very much for the research.

I would like to contribute four things for discussion:

First, ...


Dan,

Thanks for your input.

I have no idea if, or how, Apple contracts with Samsung on 14nm.

However, by simple common sense, Apple must pick the best performing and yielding process for its next flagship SoCs. It cannot afford to stumble, especially when it is now the world’s most valuable, and most talked-about, tech company.

I don’t think Apple will lock itself, by contract, into an inferior process from Samsung, a fierce competitor and critic to Apple.

The result of early simulation, even if true, is irrelevant now, because, for some time, various test chips have been built and available for evaluations.

P.S.
Second sourcing is pretty common among other companies. QCOM, AMD, MediaTek, Altera, etc., all have second, sometime third, production sources.
 
[/CENTER]

P.S.
Second sourcing is pretty common among other companies. QCOM, AMD, MediaTek, Altera, etc., all have second, sometime third, production sources.

Multi-sourcing was very common for fabless companies up until 28nm. QCOM used four different foundries at 40nm for example using the same design database (GDSII). Today it is much harder to multi-source since you cannot use the same GDSII. FPGA vendors (Xilinx and Altera) are generally single source. Xilinx was partnered with UMC until moving to TSMC at 28nm. Altera was partnered with TSMC until moving to Intel at 14nm. AMD manufactured their own CPU silicon until spinning it out to GF. The AMD GPU came from the ATI acquisition, ATI was exclusive to TSMC and still is as far as I know.

But again, Apple is not your traditional fabless companies and has not multi-sourced their SoC as of yet. Maybe at 14nm but then again maybe not.
 
As I pointed out in a prior post, linked below, it is really a cheap rumor, created by a Korean site, that Samsung has won A9 orders and started the production. The US media, however, reported this rumor as a fact and resulted in the false perception.

Samsung misinformation
https://www.semiwiki.com/forum/f293/misinformation-intel-tsmc-4952-3.html#post17455

The Taiwan news story of Dec 26 at least cited a prominent research institute MIC, instead of the unnamed “insiders” in the Korean fabrication.

TSMC to secure most orders for Apple's A9 processor: MIC | Economics | FOCUS TAIWAN - CNA ENGLISH NEWS

A story of Dec 27 from Taipei Times provides more details:

The key for TSMC to win over its main competitor, Samsung Electronics Co, in vying for Apple’s orders is its production yield, the Market Intelligence and Consulting Institute (MIC) told a press conference on technology industry trends for next year.

TSMC likely to stay key Apple supplier - Taipei Times

MIC web site includes a rich collection of its past researches. AND, the site lists the phone numbers at the bottom of each page. Journalists can call MIC for clarifications or comments. But, perhaps nowadays the US journalism has a lot to be desired.

Market Intelligence & Consulting Institute (MIC) in Taiwan
MIC- Insight with an Asian Perspective
 
Back
Top