Breaking News
News & Analysis

FinFETs Flow at Samsung, TSMC

Apple, Qualcomm are key customers
2/3/2016 10:00 AM EST
3 comments
NO RATINGS
More Related Links
View Comments: Newest First | Oldest First | Threaded View
HJ88
User Rank
Author
The Challenge is Getting Low Cost
HJ88   2/3/2016 3:09:04 PM
Many companies have initiated designs in 16/14nm FF, but there are only 3 products in high volume production at the foundry vendors.

The key challenge is getting high yields which gives low transistor costs, which is mandatory for many of the high volume applications.

16/14nm FF gives low power and high performance, but the challenge is "how to get low cost".

At 10nm, costs per design will be $200M+, with revenues needing to be 10X design costs, ie, $2B. There will not be many designs globally that will support this economic hurdle.

Mark1980
User Rank
Author
Re: a lot more
Mark1980   2/3/2016 1:56:23 PM
@realjjj Good analysis but you missed Altera - Altera have not been pulling its weight, and Intel paid Billions that could come back to bite it really hard. Altera could make or break Intel very quickly and is a BIG gamble Intel took. I dont think its going to go well for Intel.. We will wait if profits get worse every quarter for Altera and Intel.

 

TSMC are also in trouble if 16nm do not hit it out of the park. They have lost not only Qualcomm but AMD and some others also. Apple well may also give some volume to Samsung if TSMC are late or volume is bad on 16nm - its an unkown yet what will happen. But we can say Samsung are ahead with their 14nm at the moment. If the same is true for 10nm, then TSMC will take a big dive down with its stocks. But first we wait and see 16nm vs 14nm outcome.

realjjj
User Rank
CEO
a lot more
realjjj   2/3/2016 11:32:14 AM
NO RATINGS
AMD has some volume and in Q4 Zen makes them a lot more competitive, even if just on process. So very likely for them to grow  a lot for a couple of years. After that they need to look beyond PC.

Mediatek in the second half of 2016 has an 8xA53 on 16ff (assuming FFC) but that likely means that Spreadtrum's soon to launch 8 cores on 16ff is similar. The 2 are over 600 million units per year and these SoCs would be a large portion of that.

Huawei is increasing the proportion of their phones using Hisilicon (and btw look at Huawei's total R&D in 2015). They are gaining scale and are well positioned to grow long term.

Magic Leap might have hardware at MWC, with almost 800 million of new funding, who knows,maybe they amaze us and kickstart glasses as a new very high volume category.

Apple might have peaked last year. In tablets their revenue was bellow 2011 levels, the Mac will peak very soon and in smartphones they'll have 11-12.5% share this year (range depends on what the iphone 7 delivers). Remains to be seen if 2017 iphone sales can be up on year.Will be difficult as they are likely to deliver less and less while very cheap phones become more and more competent.

And, the most outrageous part, will Intel own it's fabs by 2020? The PC market could be close to 150 million units by then. Combined with some share loses and declining ASPs,  their PC revenue could be 3 times lower in 2020. Some share loss in server to AMD and ARM and they would really need new segments with high ASP and foundry clients or a lot of value would be lost on the foundry side as they drag it down with their declining volumes. Foldable screens in mobile and then glasses will hit the PC market hard and Intel's board should look into spinning off the foundry sooner rather than later. I know that by default , Intel without fabs is crazy but there is a high chance that Intel's revenues will decline a lot in the next decade and they drag the fabs down with them. Chances are , the board will panic when foldable screens reach reasonable prices and decent implementations causing the PC market to decline a lot faster but that would be a bit late (but not too late).

Maybe it used to be Qualcomm and Apple for a few years but that's changing.

Most Recent Comments

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed