Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/intel-rolls-14nm-broadwell-in-vegas.5189/
        )

    [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
)

Intel Rolls 14nm Broadwell in Vegas

astilo

New member
SAN FRANCISCO – Intel announced at CES 2015 the Broadwell family, its fifth-generation Core processors. The 14 new chips are essentially versions of the company's 22nm Haswell architecture made in its new 14nm process
This can be considered the real Intel 14nm mass production release date (Q1 2015).

Intel will offer dual and quad-core chips -- 10 processors at 15 Watts (both Core i5 and i7 chips) with Intel HD graphics, and four 28 Watt products with Intel Iris Graphics spanning i3, i5, and i7 lines. The dual-core chips have 1.9 billion transistors, a 35% increase over the prior generation, and a 133 mm2 footprint that is approximately 50mm2 smaller than its predecessors. The 15W chips have data rates up to 3.1 GHz while 28W i7 cores hit up to 3.4 GHz.
The Broadwell chips have L3 caches ranging from 2 to 4 Mbytes, roughly the same as Haswell.
So, no desktop CPUs.

While the new Core line has modest improvements in productivity, Intel hopes battery life gains will encourage users to buy new devices, said Karen Regis, director of notebook roadmap and strategy for Intel's PC Client Group.
Broadwell will power "a diverse range of form factors" from notebooks to so-called two-in-one tablet/notebook hybrids that are "seeing great retail sales momentum," she said. In addition, OEMs will announce the first Broadwell-based Chromebooks this month with the first systems in market in February.
An Intel spokesperson said the company expects the first Broadwell OEM systems in January 2015. Designs with Iris graphics and wireless gigabit are expected by the end of the first quarter of 2015. Prices range from $107 for 15W Celeron processors to $426 for 28W Core i7 processors.
Sure Intel hopes so.

Full article here:
Intel Rolls 14nm Broadwell in Vegas | EE Times
 
14nm may have been a tough road for Intel but they are now at volume at this node. Can anyone else say this?
 
14nm may have been a tough road for Intel but they are now at volume at this node. Can anyone else say this?

The foundries are 6 months behind them so the famed Intel process lead has decreased significantly compared to previous nodes. Intel took a much more difficult road of course as Scott Jones has articulated quite nicely:

Apples Versus Zebras


The Intel Cherry Trail announcement was much more significant for me. I have a feeling that it will be a more robust SoC than previously imagined.
 
I hope you are right about the 6 months’ time frame but outside of TSMC and possibly Samsung in Austin I just don't see it. I'd say a year is far more likely. This stuff is tough and Intel has some inherent advantages in that they don't have to fab multiple designs.
 
The dual-core chips have 1.9 billion transistors, a 35% increase over the prior generation, and a 133 mm2 footprint that is approximately 50mm2 smaller than its predecessors.
(1+35%)/(133/(133+50))=1.858
~ 85.8% density improvement within the same tech node. :confused:
During video playback, the same chip saw a 90 minute battery increase to 8.7 hours while operating at approximately 4.5W.
~ 21% power efficiency improvement. :confused:
FINFET design is really tricky, isn't it? :eek:
 
(1+35%)/(133/(133+50))=1.858
~ 85.8% density improvement within the same tech node. :confused:
Haswell dual core has 1.4 billion xtors and a 177 mm2 size (~8M/mm2), and it is 22nm.
Broadwell dual core has 1.9 billion xtors and a 133 mm2 size (~14.3M/mm2), and it is 14nm.
14.3/8 ~1.8, so yes a 80% count increase, but node to node.
To be clear, the equivalent 32nm Sandybridge dual core has a 4.2M/mm2 density, so the jump from 32nm to 22nm was 1.9 or a 90%.
Apple A7, 1B xtors on 102 mm2 at 28nm (9.8M/mm2), Apple A8, 2B xtors on 89 mm2 at 20nm (22.5M/mm2) --> ~2.3 or a 130% increase node to node (although not from the same foundry).
As said, Intel fans prefer another metric to assess the density. We have several posts about that, so no reason to repeat things.

~ 21% power efficiency improvement. :confused:
FINFET design is really tricky, isn't it? :eek:
22nm Haswell was already a FinFET device. The gain is more optimization and shrinkage related.
 
;)Jumping concept too fast. You claimed real 14nm just launched in Jan. So core M was categorized into Haswell. (My bad joke, obviously very confusing and does not work.) The chip density & GPxM1P measures correlation certainly represents some design complexity. I will make metric comparison.
 
Back
Top