Samsung and TSMC to Share Production of Apple's 14-nm A9 Chips in 2015

[MisakixMikasa]

I think that the point being expressed is that ARM will be "good enough" for many traditional desktop applications; it doesn't imply that ARM will supplant Intel on the desktop. My opinion is that Apple will at some point build an ARM powered desktop slotted in price below the Mac Mini, as well as a Mac Book Air class notebook powered by ARM that might be an iPad and separable keypad.

Regrettably for us all, Apple (desktop) is joined at the hip to a certain extent with Microsoft as so many users still require Windows only applications on Macs.

Probably no. If Apple to built ARM based Mac, that means we get intel based Mac and ARM based Mac at same time, that also means all the application need to be written with ARM code and x86 code. Which will confuse customers like Microsoft did with their Windows 8 Pro and Windows RT.

The reason it won't happen anytime soon is that ARM is not powerful enough to emulating all x86 applications. ARM based processors will also have hard time to match the x86 version of application in terms of performance. That could be deal breaker for most people.

 

[Traverse]

Regrettably for us all, Apple (desktop) is joined at the hip to a certain extent with Microsoft as so many users still require Windows only applications on Macs.

This was my first thought, and biggest fear, when I heard speculation on Apple designed processors for future Macs. I need to run some Microsoft only software for my degree. It doesn't matter how fast Safari is or Pages runs if I can't use the software I need. Though it is regrettable....

 

[bigwig]

Does this mean the iPhone 6 will have an A8 rather than a slightly improved A7?

 

[TMay]

How is Apple in denial. They never said they were going to remove any reliance on Samsung. They never said they were cutting all ties. They never said they would never use Samsung parts.

What are they denying?

Agreed

The only friction Apple has with Samsung is the parts that has been decided in court and/or on its way through the appeal process.

The part about "less reliance on Samsung" is just the common practice of multi sourcing production. Apple can't risk single sourcing key production, whether processors, screens or memory. Samsung has been a very good partner, and Apple an excellent teacher.

I don't see much change in the mobile duopoly of Apple/Samsung anytime soon.

----------

Probably no. If Apple to built ARM based Mac, that means we get intel based Mac and ARM based Mac at same time, that also means all the application need to be written with ARM code and x86 code. Which will confuse customers like Microsoft did with their Windows 8 Pro and Windows RT.

The reason it won't happen anytime soon is that ARM is not powerful enough to emulating all x86 applications. ARM based processors will also have hard time to match the x86 version of application in terms of performance. That could be deal breaker for most people.

I should have added in there that these will be iOS only devices, not hybrids, and I am in complete agreement with you.

 

[Gudi]

A9 maybe close to intel processor from three or five years ago, but by then, Intel processor will be even more powerful than current generation.

Doesn't matter. Even five years ago Intel chips were deemed to be too powerful for 90% of all tasks and Netbooks with even older chips emerged as a growing category. In mobile it's all about heat, battery life and price.

The simple reason that everyone can build their own ARM chips without paying Intel, might marginalize x86 in a few years. Much like with Windows, it's mostly the support of legacy software that keeps the Intel monopoly alive.

 

[MikhailT]

First thing to point out is that this is from Digitimes, a horrible prediction center. They constantly get stuff wrong, with a lucky guess every once in a while.

In my opinion, this isn't A9 at all (guessed by Digitimes most likely), and I don't think it'll be ready at all in 2015. They'd start producing some chips in 2015 on 14nm for validation process with Apple and if everything looks good, they'd switch over to mass production in the 4th quarter for release in first half of 2016.

We're likely to stay in the 2x nm range for the next two years unless Samsung/TSMC figured out how to quickly upgrade to 14nm fabs quickly without any issues and cheaply, which I doubt.

 

[commander.data]

We're likely to stay in the 2x nm range for the next two years unless Samsung/TSMC figured out how to quickly upgrade to 14nm fabs quickly without any issues and cheaply, which I doubt.

http://www.eetimes.com/document.asp?doc_id=1319679&page_number=4

It's quite likely TSMC's claims of moving from 20nm to 16nm in 1 year are accurate. The 16nm process is actually not a fully new process and uses the same back end as the 20nm process. It's just applying FinFET 3D transistors to the 20nm process which is why despite the name, the 16nm process won't actually offer much smaller transistors. They are only claiming a 5-10% improvement in die area. The advantages of the 16nm process over the 20nm process are just in performance and power.

http://www.eetimes.com/document.asp?doc_id=1319679&page_number=2

They actually have announced more design wins for the 16nm process than the 20nm process, which means many customers are confident the 16nm process will be delivered on time and are willing to skip the 20nm process.

 

[Hihowareyou]

Does this mean the iPhone 6 will have an A8 rather than a slightly improved A7?

That's what I'd assume.
I'm excited about possible quad-core A9 as well. =)

 

[techwhiz]

Probably no. If Apple to built ARM based Mac, that means we get intel based Mac and ARM based Mac at same time, that also means all the application need to be written with ARM code and x86 code. Which will confuse customers like Microsoft did with their Windows 8 Pro and Windows RT.

The reason it won't happen anytime soon is that ARM is not powerful enough to emulating all x86 applications. ARM based processors will also have hard time to match the x86 version of application in terms of performance. That could be deal breaker for most people.

Actually it wouldn't be two pieces of code.
MacOX is for the most part agnostic and ran on PPC and Intel at the same time.
There is a thing called a "fat binary".

On the other hand, ARM just can't keep up with Intel in the desktop arena.

 

[yellowtruck]

Only on forums. Elsewhere it's just called smart business decisions.

‘mmm, I'm not so sure "smart" is the best word, more like, no other option as far as cost and volume capacity are concerned. The “smart” thing would never have any of your technology blueprints in the hands of your fierce competitors who then are gaining the knowledge of its design technological advances and operations by assembling it in mass production. Like Pepsi brewing cokes’ recipe and bottling it for them. Might save cost in the moment but as we have seen Samsung corporate culture has been built on copying technology and building upon it and making it cheaper. That’s why so many American TV companies no longer exist like Phillips and Zenith. They were ripped off then undercut by cheaper Asian market electronics. It would be smarter to quickly sever all ties with Samsung and most other Asian markets. Otherwise, we will lose tech innovation all together with Apple and… are there any other US hardware tech companies anymore?

 

[Menel]

Does this mean the iPhone 6 will have an A8 rather than a slightly improved A7?

Why wouldn't it?

iPhone 4: A4 (single CortexA8)
4S: A5 (dual CortexA9)
5: A6 (dual Swift; custom but with some parallels to CortexA15)
5S: A7 (dual Cyclone)

6: would clearly have an A8

 

[MisakixMikasa]

Actually it wouldn't be two pieces of code.
MacOX is for the most part agnostic and ran on PPC and Intel at the same time.
There is a thing called a "fat binary".

On the other hand, ARM just can't keep up with Intel in the desktop arena.

Even with universal binary, there still require two piece of installer. You cannot simply load x86 code to ARM.

And I should add, Intel based Mac can emulate PPC application with emulator built in OS X up until Snow Leopard I believe. Once Apple get ride of that with Lion, you simply cannot run PPC app on Intel based Mac anymore. And there is reason you cannot run Intel base application on PowerPC Mac.

Either way, when you need run existence x86 application without recompile, you need emulation. Emulation require lots of computational power and ARM simpler does not have.

If Apple choose to go for ARM Mac, they are still in line with Microsoft's Windows. Microsoft chose not to provide emulation because lack of power with ARM processors.

----------

Doesn't matter. Even five years ago Intel chips were deemed to be too powerful for 90% of all tasks and Netbooks with even older chips emerged as a growing category. In mobile it's all about heat, battery life and price.

The simple reason that everyone can build their own ARM chips without paying Intel, might marginalize x86 in a few years. Much like with Windows, it's mostly the support of legacy software that keeps the Intel monopoly alive.

Still does not matter. Not matter what, Apple cannot compete with Wintel base with price. Even Apple release low cost ARM base Mac, the price will likely match Wintel box. However, he performance will not even come close to Wintel box or Mac Mini.

Also, if you want marginalize x86, you simple cannot do without Microsoft and Apple abandoning x86 at same time and increase ARM performance at same time. At same time, make everybody transition to ARM is highly unlikely.

Here is reason why ARM simply cannot enter high performance club and consumer based computer for many many years.

 

[D.T.]

14-nm is pretty astounding, can’t wait to see the performance/battery, final packaging with GPU, etc. We’re at 28 with the A7, assuming there’s a decrease planned for ’14 in the 22-ish range for the iPhone 6 and iPad Air 2

Side note: I’ll swear there are some users on here that go directly to the search, and enter ‘samsung’ or ‘google privacy’ every_single_day

 

[MacLC]

You fail to realized that A7's benchmark is close to 2010 Mac Mini. Mac mini never used desktop grade CPU, all the Intel Mac Mini used mobile Intel CPU.

No way, why? The A7 chip is just merely equivalently to my 2005 AMD Athlon X2 with 1GB of RAM running light weighted xUbuntu. I have done the benchmark before. A7 does not even outperform my old Intel Pentium Dual Core. Most people in this days have at least some short of Dual core desktop processor inside their computer. A7 is far away from Intel Core i3.

A9 maybe close to intel processor from three or five years ago, but by them, Intel processor will be even more powerful than current generation.

The performance gap between ARM and x86 is huge. A7 might able to do one heavy application at time with no problem, but it will struggle when multitasking between multiple application running at same time. I could run games, AutoCAD, Photoshop at same time on my desktop setup, but no way with iPad.

You may be right on one level, but isn't being such a good news bear a little hard?

Intel still makes Pentiums in the form of non-hyperthreaded i3 CPUs that the A7 can compete with. There are also low power all in one desktops that use Atom CPUs that the A7 can compete with.

Humor my comparison of the A7 to CPU in '10 Mac Mini and '13 iMac
Performance (higher is better) of compared to A7
A7:100%; '10 Mini: 167%; '13 iMac 333% ('13 iMace)
Power Draw (lower is better)
A7: 100%; '10 Mini: 555%; '13 iMac: 1444%
Cost (lower is better):
A7: 100%; '10 Mini: 1050%; '13 iMac: 960%

I beg to disagree on the A7's weak multithreading but that's another issue. A dual CPU (8 core) with 4-8GB of RAM would compete well with the Macbook Air and at less power draw and much less price. The price of the MBA could easily drop by $200-300 and Apple would still make more money on them than on the iPA or MBA today.

The A7 is just as fast if not faster than my iMac's T7700 which is fast enough for me. The only limitation is iOS. I can run Snow Leopard like a dream and kind of run Mavericks and surely could the iPA as well. I used to laugh at ARM but recently I've become a fan. I just wish Apple would let OSX run on ARM. ARM is essentially RISC as was PowerPC that Apple used for 11 years until switching to Intel/CISC. Re-releasing OSX on RISC would not exactly be rocket science.

 

[smulji]

Apple has claimed the A7 is now a Desktop-class 64-bit processor (and benchmarks show that it does perform as well as a 2010-era Desktop CPU) - this poses some interesting things for upcoming chips.
Just picture the type of advanced by the time the A9 shows up. The performance and lower power requirements. It would probably be a better performer than what most people have in their computer.

Having two separate companies (TMSC and Samsung) both doing "part" of the work for iOS devices may not be the end-goal. What if the A9 out-performs Intel equivalents, and Apple decides they want them in their Laptop and Desktop systems?

Apple could request TMSC and Samsung both ramp up to 100% each to greatly increase the amount of chips manufactured, just to be able to get them in all iOS and OS X devices.

"What if the A9 out-performs Intel equivalents, and Apple decides they want them in their Laptop and Desktop systems?"

That's a matter of when, not if. Apple likes the control the entire widget. They will.

----------

You fail to realized that A7's benchmark is close to 2010 Mac Mini. Mac mini never used desktop grade CPU, all the Intel Mac Mini used mobile Intel CPU.

No way, why? The A7 chip is just merely equivalently to my 2005 AMD Athlon X2 with 1GB of RAM running light weighted xUbuntu. I have done the benchmark before. A7 does not even outperform my old Intel Pentium Dual Core. Most people in this days have at least some short of Dual core desktop processor inside their computer. A7 is far away from Intel Core i3.

A9 maybe close to intel processor from three or five years ago, but by them, Intel processor will be even more powerful than current generation.

The performance gap between ARM and x86 is huge. A7 might able to do one heavy application at time with no problem, but it will struggle when multitasking between multiple application running at same time. I could run games, AutoCAD, Photoshop at same time on my desktop setup, but no way with iPad.

"The performance gap between ARM and x86 is huge."

CURRENTLY, that is true. You're assuming that will always be the case. For all we know, in 2 to 4 years, that gap might not exist.

 

[winston1236]

‘mmm, I'm not so sure "smart" is the best word, more like, no other option as far as cost and volume capacity are concerned. The “smart” thing would never have any of your technology blueprints in the hands of your fierce competitors who then are gaining the knowledge of its design technological advances and operations by assembling it in mass production. Like Pepsi brewing cokes’ recipe and bottling it for them. Might save cost in the moment but as we have seen Samsung corporate culture has been built on copying technology and building upon it and making it cheaper. That’s why so many American TV companies no longer exist like Phillips and Zenith. They were ripped off then undercut by cheaper Asian market electronics. It would be smarter to quickly sever all ties with Samsung and most other Asian markets. Otherwise, we will lose tech innovation all together with Apple and… are there any other US hardware tech companies anymore?

No comparison. Coke and Pepsi are both "bottled" by a vast array of independent companies across the world. Phillips is a Dutch company based in Amsterdam btw, not American.

 

[Tiger8]

Yep it's been reported that Apple has always had a great relationship with the manufacturing side if Samsung.

And why wouldn't they? Love them or hate them Samsung products are top notch, and Apple is a huge customer for Samsung; it's a mutually beneficial billions of dollars worth relationship.

I know Apple is venturing out and diversifying their suppliers, but still, they will be their biggest supplier at least for the next few years.

 

[alent1234]

"What if the A9 out-performs Intel equivalents, and Apple decides they want them in their Laptop and Desktop systems?"

That's a matter of when, not if. Apple likes the control the entire widget. They will.

----------



"The performance gap between ARM and x86 is huge."

CURRENTLY, that is true. You're assuming that will always be the case. For all we know, in 2 to 4 years, that gap might not exist.

x86 has larger cache. until an ARM chip has the same it won't be faster

 

[Tiger8]

That’s why so many American TV companies no longer exist like Phillips and Zenith. They were ripped off then undercut by cheaper Asian market electronics. It would be smarter to quickly sever all ties with Samsung and most other Asian markets.

First of all, Philips is not an American company.

Second, do your research about 'American TV Companies'. You see in a global economy, the best products seem to win, and the reason why 'American TV Companies' lost to others is the dramatic drop in quality; American consumers are smart, if they buy a TV that has crappy picture and lasts only a couple of years, they'll switch brands. And for the record, SHARP (Japanese) were the major disruptor back in CRT days, and they never were cheaper TVs, but they offered fantastic quality and products that last forever.

But just like Sharp and other Japanese brands owned the market in the late 70s to late 90s, SAMSUNG came with the LCD revolution and producted the best TVs in the world, and now they are #1, worldwide.

 

[samcraig]

‘mmm, I'm not so sure "smart" is the best word, more like, no other option as far as cost and volume capacity are concerned. The “smart” thing would never have any of your technology blueprints in the hands of your fierce competitors who then are gaining the knowledge of its design technological advances and operations by assembling it in mass production. Like Pepsi brewing cokes’ recipe and bottling it for them. Might save cost in the moment but as we have seen Samsung corporate culture has been built on copying technology and building upon it and making it cheaper. That’s why so many American TV companies no longer exist like Phillips and Zenith. They were ripped off then undercut by cheaper Asian market electronics. It would be smarter to quickly sever all ties with Samsung and most other Asian markets. Otherwise, we will lose tech innovation all together with Apple and… are there any other US hardware tech companies anymore?

Not a valid comparison. It would be more akin to Coke using Pepsi's glass fabrication. Or their sugar refinery. The chip being manufactured by Samsung is not the iPhone. Or iPad. It's a part. An important part. But not at all the entire product.

No comparison. Coke and Pepsi are both "bottled" by a vast array of independent companies across the world. Phillips is a Dutch company based in Amsterdam btw, not American.

exactly

 

[Gudi]

Not matter what, Apple cannot compete with Wintel base with price.

They could, but they choose not to. Apple wants to compete on better experience. A super-light, fan-less design with incredibly long battery life could completely redefine the MacBook Air category and all its clones.

Even Apple release low cost ARM base Mac, the price will likely match Wintel box. However, the performance will not even come close to Wintel box or Mac Mini.

A lower cost cpu is a necessity for all manufacturers to increase margins, no matter the retail price. And ARM will come to laptops first, not desktops. Which is the bigger half of the PC market anyway. Despite the x86 chips in the 13-inch MacBook Pro's being only dual-core ULVs, they still run hot when used at 100% and can eat up all the battery in just 2 hours. So all OSX Mavericks tries to do, is send these stupid CPUs to sleep. That is simply not feasible for a mobile device.

Also, if you want marginalize x86, you simple cannot do without Microsoft and Apple abandoning x86 at same time and increase ARM performance at same time.

Neither of that is necessary. The PC market is already shrinking and tablets and smartphones are overtaking. The benefits of ARM architecture, less heat and lower power draw are equally important for all mobile devices tablets and laptops alike.

At same time, make everybody transition to ARM is highly unlikely.

It is highly unlikely that the Intel monopoly will not be challenged by manufacturers, if there is any possibility.

Here is reason why ARM simply cannot enter high performance club and consumer based computer for many many years.

Tablets are also consumer computers. And real high performance computers (like the Mac Pro) are a niche market. The new generation of gaming consoles are a higher performance consumer market and they have all chosen AMD chips because of the lower price point.

 

[mpantone]

Does this mean the iPhone 6 will have an A8 rather than a slightly improved A7?

Maybe, maybe not.

My guess is that engineering samples of both CPUs will end up in prototype units somewhere in a lab in Cupertino.

At some point, the senior management will look at chip yield forecasts and decide whether or not to move with one or the other. Certainly Apple hopes to use a more capable CPU, but the part must ship in a certain quantity at a certain price by a certain date.

Those are supply chain factors that no one here on MacRumors can predict. It's up to Apple's component partners to deliver. You nor I can predict whether or not Samsung, TSMC, or anyone else will succeed for an unknown processor design.

 

[koban4max]

How is Apple in denial. They never said they were going to remove any reliance on Samsung. They never said they were cutting all ties. They never said they would never use Samsung parts.

What are they denying?

That they need each other. Come on...u forgot that samsung vs apple trial? the trial that led to somewhat divorced?

 

[MikhailT]

http://www.eetimes.com/document.asp?doc_id=1319679&page_number=4

It's quite likely TSMC's claims of moving from 20nm to 16nm in 1 year are accurate. The 16nm process is actually not a fully new process and uses the same back end as the 20nm process. It's just applying FinFET 3D transistors to the 20nm process which is why despite the name, the 16nm process won't actually offer much smaller transistors. They are only claiming a 5-10% improvement in die area. The advantages of the 16nm process over the 20nm process are just in performance and power.

http://www.eetimes.com/document.asp?doc_id=1319679&page_number=2

They actually have announced more design wins for the 16nm process than the 20nm process, which means many customers are confident the 16nm process will be delivered on time and are willing to skip the 20nm process.

But the article in the OP is talking about 14nm and both Samsung/TSMC doing that.

Would Apple really do two separate designs for both companies?

I wish both fabs all the luck, the 3d stacking seems like it's going to ramp up to 16nm and 10nm easily within the next three years.

 

[Tknull]

Business is business I guess. nothing personal.
and currently samsung can cope with apple demand to produce these chips.

I think the "soap opera" has existed far more in rumorville than in reality. These types of sites have made big todo about relationship issues/changes/etc... where no actual statement or policy has been espoused by any of the parties involved. Rumour is probably always far more dramatic than actual reality.