Blog
Comment
peter.clarke
armguru
The 64-bit extensions primarily give you an extended addressing capability and ...
London Calling: Apple's A7 expected to tape out in March
Peter Clarke
3/15/2013 8:07 AM EDT
Foundry Taiwan Semiconductor Manufacturing Co. Ltd. will finish the tape out of Apple's A7 processor in 20-nm CMOS in March in time for volume production in early 2014, according to a report from DigiTimes which referenced unnamed sources.
Completing the design in 1Q13 will allow the chip, expected to be used in forthcoming generations of iPhone and iPad, to begin so-called "risk production" in May or June and lead to high volume production in 1Q14, the report said. TSMC is adding manufacturing capacity at its Fab-14 Gigafab where is plans to make the A7 processor, the report added.
Fab-14 in Tainan, Taiwan is not only set to be TSMC's first fab to produce 20-nm CMOS system chips in volume but it also earmarked as the place where it will bring up its follow-on 16-nm FinFET manufacturing process.
That TSMC is set to make the A7 processor for Apple is no surprise. We've heard such things before that did not come to pass. Talk of Apple moving its processor production away from sole supplier Samsung has been rife since at least 2011, with a particular piquancy as Apple and Samsung have engaged in legal battles around the world over their mobile devices.
It is thought that Apple engaged with TSMC to try and get the A6 application processor out in 2011 on the 28-nm process node but that the difficulties of migrating design elements, particularly around such as issues as collateral IP may have thwarted the move.
Talk also continues to swirl around the possibility that Intel could also be a manufacturer for Apple. Intel does supply x86 processors to Apple for use in its MacBook computers and is gradually increasing its foundry engagements.
However, even though Intel is well-versed in making complex microprocessors, offering all the IP needed to let a third-party design their own processor and then make it is a highly complex design interaction and possibly one that would be a stretch for Intel's foundry operation right now. Or, at least, Apple might think so if it got its fingers burned at the 28-nm node.
Nonetheless the possibility remains that Apple could be moving to a multiple supplier strategy for its own processors. However, it seems likely that will be implemented as serial engagements.
In other words, while Samsung may go on supplying A6 processors made using 32-nm/28-nm CMOS to Apple for some time, TSMC looks set to be the incumbent at 20-nm but it may have been made clear to the foundry that Apple is already working with Intel aiming at the 14-nm and 10-nm FinFET as a potential alternative to TSMC's 16-nm FinFET roadmap.
Related links and articles:
DigiTimes article
TSMC test runs Apple processor, says report
Analyst: Apple qualifying TSMC for A6
TSMC breaks ground for FinFET fab
TSMC confirms U.S. wafer fab site hunt
Navigate to related information


MikeSmith2011
3/15/2013 9:00 PM EDT
Does this chip have the homegrown 64b ARM processor?
Sign in to Reply
peter.clarke
3/18/2013 7:01 AM EDT
@MikeSmith2011
Good question.
I don't know the answer but I suspect not.
It just seems a bit soon to be going to ARMv8 and Cortex-A57. Also ARMv8 is still seems more relevant to server-end than the client end. Although it will come into clients in time no doubt.
It also seems a bit soon as Apple has been developing its own ARMv7-compatible cores rather than taking cores off the shelf.
Still Apple has tendency to be more radical than most companies and it is to be expected that they will make use of an architectural ARMv8 license at some point.
Sign in to Reply
asic_pal
3/16/2013 4:06 PM EDT
I am sure Intel might be working on 8-10nm designs, FPGA's alreay producing 20nm products and working on 14nm starting next year. Almost all Tier1 companies must be taping out chips in 20nm this year.
But it's kind an interesting that how much people of interested to know about Apple tapeout from an unknown source, than some interesting feature facts of Samsung Galaxy IV released yesterday!
Kudos to Apple!
Sign in to Reply
HS_SemiPro
3/16/2013 4:41 PM EDT
I have been thinking the same, Why is there not a big article in eetimes about what Samsung showed in their Thursday's unveiling of new Galaxy phone. If it were apple, there will be a many articles, only thing I see a 2 paragraph saying what's missing.
Bias? or paid by apple or not paid by Samsung?
Sign in to Reply
eewiz
3/18/2013 12:51 AM EDT
this is the news about a chip release.. which is more relevant to an EE than the features of a new phone, which anyways didnt have any new radical features than incremental software features. Similar chip release news from samsung were always posted in EEtimes. For eg. http://www.eetimes.com/design/microcontroller-mcu/4404719/Samsung-reveals-eight-core-Exynos-processor
http://www.eetimes.com/electronics-news/4404991/London-Calling-Whose-GPU-in-Exynos-Octa
http://www.eetimes.com/electronics-news/4371786/Samsung-announces-quad-core-Exynos-processor
http://www.eetimes.com/electronics-news/4407832/Slideshow--Mobile-at-ISSCC-with-Samsung--Mediatek
Sign in to Reply
Big Paul
3/18/2013 12:55 PM EDT
EETimes is not targeted at people looking for smart phone reviews. If that is what you want, read any of the hundreds of articles at places like Ars Technica or CNET.
Repeat after me: there is no global Apple conspiracy. An article about a processor taping out is not a global conspiracy. Not mentioning Samsung in every article is not a global conspiracy.
eewiz conveniently identified plenty of EETimes Samsung articles to demonstrate there is no global conspiracy.
Sign in to Reply
melgross
3/19/2013 9:32 AM EDT
It would seem that 2014 is too late for a new Apple design. If the rumors are false, we will again see a new phone at the latest, in September or October. If they are true, and a new phone is out earlier, then the timing is even worse, impossible really.
I can't imagine Apple going without a new chip for the new phone and iPad 5. So either they are still using Samsung for the part, or something is happening that isn't being reported.
Sign in to Reply
armguru
3/19/2013 12:02 PM EDT
The 64-bit extensions primarily give you an extended addressing capability and presumably there is some price to pay for this in power consumption - moving around 64-bit addresses costs more energy and also worsens cache hit rates than working with 32-bit addresses. Unless you need the address space why would a mobile design move to the 64-bit architecture?
Sign in to Reply
peter.clarke
3/19/2013 2:16 PM EDT
@armguru
The ARMv8 architecture fully supports 32-bit addressing but allows 64-bit addressing in a special mode.
I remember ARM CTO Mike Muller saying back in 2010 that clients are going to have to become more like servers and take on 64-bit addressing. Obviously the amount of physical memory continues to increase.
Some applications like face recognition within video streams might be candidates for 64-bit addressing on mobiles?
If anyone can think of others please chime in.
Sign in to Reply