Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
63,447
30,653



Samsung and Globalfoundries announced yesterday that the two companies will adopt the same chip production process as they upgrade their manufacturing facilities in preparation for the next generation mobile devices. (Via Wall Street Journal) This announcement adds credence to an earlier rumor that Apple was negotiating with GlobalFoundries to start making the A-Series chip in the company's Malta, New York Fab 8 facility, with Samsung helping in the early stages of production.

global-foundaries-logo.png
As part of the agreement, Globalfoundries has agreed to drop its current 14 nanometer circuit technology and will instead license Samsung's 14 nanometer process. This standardization of production processes could have a significant impact on Apple, which can now source its A-series processors from either Samsung or GlobalFoundries without any additional engineering effort.
"What really drove us to this agreement was a strong pull from customers," said Ana Hunter, Globalfoundries' vice president of product management. "They see the value in really having one process where they have choice and flexibility in the manufacturing options that they have."
Intel is the first company to start producing chips based on 14-nanometer technology with production starting in the current quarter. Apple supplier Samsung is expected to begin volume production of its 14 nanometer chips in late 2014, while Globalfoundries may introduce the process in early 2015.

There is no information on how this agreement affects Apple's current chip manufacturing deal with Taiwan Semiconductor Manufacturing (TSMC). Apple reportedly contracted with TSMC to start producing its A-series chips in early 2014 using 20-nanometer technology. Apple turned to TSMC for its chip manufacturing needs as the Cupertino company strives to lessen its reliance on Samsung to meet its production needs. This balance of suppliers helps alleviate supply constraints that result from production problems at a single supplier.


Article Link: Apple Could Benefit as Samsung and Globalfoundries Standardize Mobile Chip Production
 

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
http://semiaccurate.com/2014/04/17/semiaccurate-right-apples-foundry-plans/
http://semiaccurate.com/2014/04/17/global-foundries-samsung-sync-14nm-processes/

That means, that in next year we will see Chips based on this Process. And it will include GPU made by Apple in the same manner as Cyclone cores in A7 are...
http://semiaccurate.com/2014/03/25/details-apples-gpu-emerge/

http://semiaccurate.com/2013/03/25/what-is-apple-doing-at-14nm/ as a bonus... ;)


A8 - 20nm, quad core Cyclone 1.4/1.5 GHz CPU, PowerVR G6650?
A9 - 14 nm, Quad Core Hurricane(?) 1.5 GHz CPU, Apple Vision GP9(Based on Kepler/Maxwell Architecture from Nvidia)?
 
Last edited:

chrmjenkins

macrumors 603
Oct 29, 2007
5,325
158
MD
http://semiaccurate.com/2014/04/17/semiaccurate-right-apples-foundry-plans/
http://semiaccurate.com/2014/04/17/global-foundries-samsung-sync-14nm-processes/

That means, that in next year we will see Chips based on this Process. And it will include GPU made by Apple in the same manner as Cyclone cores in A7 are...
http://semiaccurate.com/2014/03/25/details-apples-gpu-emerge/

http://semiaccurate.com/2013/03/25/what-is-apple-doing-at-14nm/ as a bonus... ;)


A8 - 20nm, quad core Cyclone 1.4/1.5 GHz CPU, PowerVR G6650?
A9 - 14 nm, Quad Core Hurricane(?) 1.5 GHz CPU, Apple Vision GP9(Based on Kepler/Maxwell Architecture from Nvidia)?

Semiaccurate is a rumors site full of speculation that doesn't turn out to be correct a fair amount of the time.

For instance, they claimed the Xbox One APU taped out late 2011 and had a PowerPC core: http://semiaccurate.com/2011/12/05/exclusive-xbox-next-chip-just-taped-out/

Charlie may have sources, but it's a lot of guesses.

And it's unlikely Apple chooses to go Nvidia. PowerVR is optimized for mobile and better at a perf/W perspective.
 

paulrbeers

macrumors 68040
Dec 17, 2009
3,963
123
Yes, it's not unreasonable to think that Apple has extensive access to their IP for potential custom implementations.

And since they own a significant piece of powerVR, they can also push their own agendas on them (i.e. what Apple wants, Apple will get). No way Apple goes to Nvidia for GPU's. If anything, Apple would consume the other 90% of PowerVR (i.e. purchase) and use their designs as their custom GPU's going forward if they felt like they weren't getting what they needed out of PowerVR.


No way apple jumps ship and goes to Nvidia....
 

koyoot

macrumors 603
Jun 5, 2012
5,939
1,853
What I meant, was that GPU from Apple would have few ideas of design taken from "bigger" players, in the same way Apple took ideas for Cyclone cores from Intel.

Obviously, it would be PowerVR Architecture. But Enhanced by ideas from Nvidia/AMD.

If they could jump in performance, by licensing the architecture, not simply the cores, and designing them from ground up by them, and getting it to levels, that noone could meet - that would be an astonishing achievement.
 

chrmjenkins

macrumors 603
Oct 29, 2007
5,325
158
MD
What I meant, was that GPU from Apple would have few ideas of design taken from "bigger" players, in the same way Apple took ideas for Cyclone cores from Intel.

Obviously, it would be PowerVR Architecture. But Enhanced by ideas from Nvidia/AMD.

If they could jump in performance, by licensing the architecture, not simply the cores, and designing them from ground up by them, and getting it to levels, that noone could meet - that would be an astonishing achievement.

Cyclone takes ideas from all the big wide desktop designs that have been prevalent over the past decade, whether it be Intel, AMD or IBM. There's no publicly disclosed feature of it that favors any one of those three over the other.

Nvidia is using their desktop design in mobile to unify their product line and simplify development costs. It's not as optimized for mobile workloads as PowerVR parts are. They would need an astounding reason to break backwards compatibility and do custom design of a GPU of a different architecture.
 

wizard

macrumors 68040
May 29, 2003
3,854
571
What I meant, was that GPU from Apple would have few ideas of design taken from "bigger" players, in the same way Apple took ideas for Cyclone cores from Intel.
The ideas embedded in Cyclone didn't come from Intel though Intel may be the first to mass market CPUs with these sort of features. Much of what we see in a Microprocessor these days is just recycled from the Mini and Mainframe worlds of the past. The implementations may be different to avoid patent issues, for things still under patent, but the ideas are more or less the same. Also don't forget that the academic world has played an important part in the development of CPU architectures.
Obviously, it would be PowerVR Architecture. But Enhanced by ideas from Nvidia/AMD.
If Apple does its own GPU all bets are off as far as where it comes from and what is added. Over the years they have acquired companies and technologies directly related to the manufacture of GPUs. I wouldn't be surprised to find Apple working on highly integrated GPU/CPU functionality similar to AMDs HSA tech. Because of this I can also see far more versatile execution units in the GPU to better support OpenCL and the specific needs Apple has. It should be noted that Apple hasn't supported OpenCL on iOS devices yet and I suspect that is because there are shortcomings in Imaginations design.
If they could jump in performance, by licensing the architecture, not simply the cores, and designing them from ground up by them, and getting it to levels, that noone could meet - that would be an astonishing achievement.

If Apple delivers 14 nm hardware in 2015 that would be a huge achievement all by itself. I wouldn't be surprised to find out that we might see limited 14 nm hardware this year in an iWatch.

As for GPU performance I wish it was that easy! I can see Apple struggling with this if the design is totally DIY and does not borrow technology from anybody else. Let's face it we really only have a couple of independent GPU manufactures these days and only two GPU manufactures delivering high performance.
 

mdriftmeyer

macrumors 68040
Feb 2, 2004
3,809
1,985
Pacific Northwest
Developing FinFET jointly between Samsung, TSMC and GlobalFoundries has been known for a long time, but joint distribution of a unified solution is BRAND NEW to the Industry.

Besides the info on GloFo site, this technical article from Semimd gives one more insight:

http://semimd.com/blog/2014/04/17/globalfoundries-and-samsung-join-forces-on-14nm-finfets/

One of the key take aways is they are bypassing 20nm and going straight to 14nm.

----------

The standard applies to all Global Foundries fab around the world? Or just NY?

Just Malta, the most advanced one.

http://globalfoundries.com/manufacturing/300mm-manufacturing

Dresden which has a higher capacity, for older 45nm and below manufacturing isn't focused like Malta which is targeted for 28nm and below.

Fab 8 is where all the joint development is taking place, including the school of nanontechnologies.

The campus is in the middle of a $40 billion development, with Fab 8 fully operational now.

What is clearly not in the picture is TSMC, the other partner who developed FinFET 14nm with Samsung, IBM and GlobalFoundries.

IBM is getting out of the Fab business, so they will be pushing their demands off onto their partners to manufacture, while they focus on R&D.
 

wikiverse

macrumors 6502a
Sep 13, 2012
689
952
From a non-nerd perspective, this is good news for Samsung... Everyone licensing their processes means they can easily pick up overflow work and can control the development of the technology.

This doesn't so much reduce apple's reliance on Samsung, but actually increases samsung's power in chip development. So indirectly, apple just became more reliant on Samsung.
 

mdriftmeyer

macrumors 68040
Feb 2, 2004
3,809
1,985
Pacific Northwest
From a non-nerd perspective, this is good news for Samsung... Everyone licensing their processes means they can easily pick up overflow work and can control the development of the technology.

This doesn't so much reduce apple's reliance on Samsung, but actually increases samsung's power in chip development. So indirectly, apple just became more reliant on Samsung.

It's not Samsung's Processes being licensed. It's a joint cross-licensing agreement to be a first of its kind unified platform where the client can expect both Samsung and GlobalFoundries to stamp out the same solutions, reducing costs by hundreds of millions to the clients.

It's a collaboration that extends their already developed research between Samsung, Global Foundries, TSMC and IBM.

The difference is that Samsung and GF are working together because TSMC is king and IBM is out of the fab business.
 

xVeinx

macrumors 6502
Oct 9, 2006
361
0
California
Apple has hired an entire team + away from AMD's GPU division, and they have a heavy investment in openCL. They'd have to license IP from both Nvidia and AMD among others, but I would guess that you'll see them do something newish; more powerful than PowerVR, but less so than AMD or Nvidia GPUs. They have the design team and talent to do it. It's just a matter of if and when they see it as appropriate to introduce the new tech...
 

hleewell

macrumors 6502a
Oct 22, 2009
544
62
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.