Conflicting reports claim TSMC, Samsung each have upper hand for building Apple's 'A9' chip
The outlook remains murky for Apple's anticipated "A9" processor, expected to arrive in its 2015 iPhone, with a pair of new reports disagreeing on whether Samsung or TSMC will build the bulk of next-generation chips.
Citing sources within the chip fabrication industry, Srini Sundarajan of Summit Research said in a note to investors on Tuesday that he's heard Samsung is currently the frontrunner in alleged "A9" production. In the note, a copy of which was provided to AppleInsider, Sundarajan said that Samsung "has the edge" based on early testing, but added that "samples are different from production."
He also believes that chipmaker Global Foundries will be "involved" in building silicon for Apple. But he noted that the iPhone maker's interest in bringing manufacturing to the U.S., along with the GT Advanced Technologies sapphire debacle, make Global Foundries unlikely to "play a spoilsport" to Samsung and TSMC.
Sundarajan's claims align with recent reports alleging that Samsung is likely to build the majority of Apple's 2015 A-series chips. That would be a change from the current A8 processors found in the latest iPhones and iPad Air 2, which are said to be primarily built by Taiwan Semiconductor Manufacturing Co.
However, going against the grain is Market Intelligence and Consulting Institute, whose director Chris Hung was quoted by the Taipei Times this weekend as saying that TSMC is likely to remain Apple's primary supplier for the "A9" CPU. According to Hung, it's actually TSMC, and not Samsung, that has the higher yield rates in early "A9" chip production.
Still, Hung said that Samsung is still likely to receive some of the orders for Apple's anticipated "A9," as Apple prefers to spread risk to avoid potential supply issues that could hamper the massive global roll-out of its next-generation devices.
Earlier reports had claimed that Samsung was producing "A9" chips with its 14-nanometer FinFET technology, while TSMC was said to be vying for Apple's business with its own 16-nanometer FinFET plus process.
Prior to the launch of the iPhone 6 and iPhone 6 Plus, Samsung manufactured all of Apple's low-power ARM processors at its factory in Austin. But starting with the 20-nanometer A8 CPU in the iPhone 6 and iPhone 6 Plus, TSMC was said to have taken over the bulk of responsibility, though exactly how much is in doubt.
While initial reports claimed TSMC was building all A8 chips for Apple, more recent rumors have claimed Samsung is still supplying about 40 percent of the latest-generation 64-bit mobile processors. The A8 is built through a 20nm process, which itself was a reduction from the previous A7 processor.
Smaller processors are more efficient, resulting in power savings that can allow devices like the iPhone to see improved battery life, even as the chips themselves become more powerful and capable. The Apple-designed A8 chip has been found through independent benchmarks to outperform the competition while still offering excellent battery life in the iPhone 6 and iPhone 6 Plus.
Citing sources within the chip fabrication industry, Srini Sundarajan of Summit Research said in a note to investors on Tuesday that he's heard Samsung is currently the frontrunner in alleged "A9" production. In the note, a copy of which was provided to AppleInsider, Sundarajan said that Samsung "has the edge" based on early testing, but added that "samples are different from production."
He also believes that chipmaker Global Foundries will be "involved" in building silicon for Apple. But he noted that the iPhone maker's interest in bringing manufacturing to the U.S., along with the GT Advanced Technologies sapphire debacle, make Global Foundries unlikely to "play a spoilsport" to Samsung and TSMC.
Sundarajan's claims align with recent reports alleging that Samsung is likely to build the majority of Apple's 2015 A-series chips. That would be a change from the current A8 processors found in the latest iPhones and iPad Air 2, which are said to be primarily built by Taiwan Semiconductor Manufacturing Co.
However, going against the grain is Market Intelligence and Consulting Institute, whose director Chris Hung was quoted by the Taipei Times this weekend as saying that TSMC is likely to remain Apple's primary supplier for the "A9" CPU. According to Hung, it's actually TSMC, and not Samsung, that has the higher yield rates in early "A9" chip production.
Still, Hung said that Samsung is still likely to receive some of the orders for Apple's anticipated "A9," as Apple prefers to spread risk to avoid potential supply issues that could hamper the massive global roll-out of its next-generation devices.
Earlier reports had claimed that Samsung was producing "A9" chips with its 14-nanometer FinFET technology, while TSMC was said to be vying for Apple's business with its own 16-nanometer FinFET plus process.
Prior to the launch of the iPhone 6 and iPhone 6 Plus, Samsung manufactured all of Apple's low-power ARM processors at its factory in Austin. But starting with the 20-nanometer A8 CPU in the iPhone 6 and iPhone 6 Plus, TSMC was said to have taken over the bulk of responsibility, though exactly how much is in doubt.
While initial reports claimed TSMC was building all A8 chips for Apple, more recent rumors have claimed Samsung is still supplying about 40 percent of the latest-generation 64-bit mobile processors. The A8 is built through a 20nm process, which itself was a reduction from the previous A7 processor.
Smaller processors are more efficient, resulting in power savings that can allow devices like the iPhone to see improved battery life, even as the chips themselves become more powerful and capable. The Apple-designed A8 chip has been found through independent benchmarks to outperform the competition while still offering excellent battery life in the iPhone 6 and iPhone 6 Plus.
Comments
I can settle this right now.
It DOESN'T MATTER if Samsung or TSMC ends up getting the whole order, or a part of the order.
All we know for sure is that the A9 will be the world's most advanced ARM processor, just like the A8 and A7 were before it. Nothing that Qualcomm, Samsung or Nvidia has will even come close.
I don't really care WHO builds it. It's still 100% Apple designed.
I can settle this right now.
It DOESN'T MATTER if Samsung or TSMC ends up getting the whole order, or a part of the order.
All we know for sure is that the A9 will be the world's most advanced ARM processor, just like the A8 and A7 were before it. Nothing that Qualcomm, Samsung or Nvidia has will even come close.
I don't really care WHO builds it. It's still 100% Apple designed.
One worries, with Samsung's reputation, that some of the technology spills over. Also, Apple needs to stop padding Samsung's pockets.
I can settle this right now.
It DOESN'T MATTER if Samsung or TSMC ends up getting the whole order, or a part of the order.
All we know for sure is that the A9 will be the world's most advanced ARM processor, just like the A8 and A7 were before it. Nothing that Qualcomm, Samsung or Nvidia has will even come close.
I don't really care WHO builds it. It's still 100% Apple designed.
Yep. As an investor I wouldn't buy stock in either of those suppliers, I'd buy more AAPL (which might be a good idea since the end-of-the-year profit takers are currently depressing the stock price).
what if apple said "drop smartphones, we'll give you the bulk of our Ax/memory chip manufacturing"...?
more recent rumors have claimed Samsung is still supplying about 40 percent of the latest-generation 64-bit mobile processors.
Oh come on, I thought we were done with that nonsense
Oh come on, I thought we were done with that nonsense
It's no nonsense. There are components that are supplied by Samsung but not designed by Apple, the A9 does not fall in that category.
One worries, with Samsung's reputation, that some of the technology spills over. Also, Apple needs to stop padding Samsung's pockets.
That would be disaster for Samsung. Samsung Mobile can copy some design elements or features from Apple, but those are often subjective as to infringement.
If Samsung Semiconductor used Apple IP in an Exynos processor, it would be very easy for Apple to determine this through both a chip teardown/analysis and also actual testing of the chip. It would also be much easier to prove in court as we're dealing with something concrete and physical.
Nobody in the world would trust Samsung Semiconductor to manufacture their designs if they were caught blatantly stealing IP from one of their customers.
The reports aren't mutually exclusive. It's likely that one of them is building the A9 for the iPhone and the other is building the A9X for the iPad, rather than both companies making the same processor for the same device.
If I were Apple and in the position of having to work with Samsung at all, I'd have TSMC build the A9 and Samsung build the A9X, since Samsung is less trustworthy and the iPad is less important than the phone.
I'd hope TSMC gets first crack then Apple use Scamsung for any shortfall. Why ..? Oh because Scamsung are scummy servants of the evil Google.
So Tim Cook's reputation for squeezing all of the padding out of Apple's outsourcing contracts is undeserved?
Um, but GlobalFoundries has a fab in the US...
Just like Samsung built a fab in the US.
The one without a fab in the US is TSMC.
The real problem for Apple is that they need 100,000 people in a factory to assemble the number of iPhones they sell, so assembly is going to occur outside the US for a while regardless of where the components are made.
Anyway, it doesn't really matter who makes the A9 in the end. Both foundries are good choices, and the paranoia about stealing IP at this level is ridiculous. Intelligence, maybe, but what good is intelligence about the current Apple chip when that can only feed back into a 12-18 month design cycle anyway, by which time Apple will have a different chip out?
It is actually TSMC 16nm that Apple will be using.
Samsung will likely be the source of A9X, or even A8X.
If Apple's A10 processor is built using a 10 nm process in 2016, that would give it a very big lead on the competition -- and may also set the stage for replacing Intel inside of Macs. If & when that occurs, it will allow Apple to whittle perhaps $100 or more off the production cost of a Mac, and even more off the retail price.
Meanwhile, a 10 nm process (and smaller) will allow Apple to optimize the trade-off between size (smaller), power (more) and time between re-charging (longer). Mobile computing is all about those things, so by dominating in processors Apple can gain advantages that can't be duplicated (at a moderate cost, within a reasonable time frame) by Android-based mobile phones.
This was in Digitimes, October 6, and mentions taping-out 10 nm designs as early as the end of 2015: "ARM and Taiwan Semiconductor Manufacturing Company (TSMC) have announced a new multi-year agreement that will deliver ARMv8-A processor IP optimized for TSMC 10nm FinFET process technology. Because of the success in scaling from 20nm SoC to 16nm FinFET, ARM and TSMC have decided to collaborate again for 10FinFET. This early pathfinding work will provide valuable learning to enable physical design IP and methodologies in support of customers to tape-out 10nm FinFET designs as early as the fourth quarter of 2015."
2016 for 10 nm is very optomistic.
Apple currently uses the so called low power processes from Samsung and TSMC. A processor suitable for a Mac would likely change to a performance optimized process, this so that the clock rate can be boosted. In between you have the Air which might be feasible on the low power processes.
It could add up to much more than $100. Intel has very fat margins on its processors. Beyond that Apple can save big bucks by precisely tailoring the chip for the machine.
Reading some of your comments, you seem relatively knowledgeable on topics related to processors. Curious to know your thoughts on this article;
http://www.t-gaap.com/2014/12/30/is-intels-time-nearly-over-with-apple
And pay close attention to "pedro's" comments.
I can settle this right now.
It DOESN'T MATTER if Samsung or TSMC ends up getting the whole order, or a part of the order.
All we know for sure is that the A9 will be the world's most advanced ARM processor, just like the A8 and A7 were before it. Nothing that Qualcomm, Samsung or Nvidia has will even come close.
I don't really care WHO builds it. It's still 100% Apple designed.
You're right it doesn't matter because any supplier will most likely be squeezed so much that profits will be small anyway.
But it's a myth that nothing comes close to the A8. I can back that up since I have both an iPhone 6 and Note 4 in my house hold. But take a look at hard facts:
www.tomshardware.com/reviews/apple-iphone-6-plus-review,3976-14.html
Single core, yes the A8 has an edge but when you goto multi-core then it slips behind. That doesn't qualify as nothing "comes close". Plus, the Exynos version (which I don't have) is even faster but harder to find.
If you think "multi-core" doesn't matter it does, we download, update, skype and do various things in the background more and more with our phones.
So what does that all mean? If Apple does anything Arm based I suspect it will be marketed as a new product line. This to keep it clearly distinguished from the Mac i86 lineup. It could take as long as five years for the customer base to buy into a line of Macs that are Arm based. The general population would wait for early adopters to take all of the arrows.
Long term Intel either needs to become far more open about customer hardware and working with customers on specific solutions or Apple will be forced to go Arm.
Pedro is highly biased and frankly doesn't seem to understand where the industry is going. Frankly I think he is full of bullshit.
Here is the problem in a nut shell, the world is moving to SoC technology. We already see this in Apples Arm line of processors. This is allowed by the much denser processes but is required to deliver things like the iPad. The thing that everyone seems to over look these days is that the SoC is effectively the Printed Circuit Board of the 1980's. The PCB is where companies put their IP back then, today though they put that IP right on the SoC.
So we have Intel that for whatever reason can't seem to manage to produce a decent SoC even at 14nm. If they don't get their act together here Apple would be forced to go their own way with Arm. Apple doesn't currently build a lot of locally sourced IP into their Macs, however if they are expected to bring some of the goodies that we see in IOS devices to their Macs then they need to get Intel to atleast put those features into their processors in a generic manner. If not that they need to get Intel to do custom.
An example here is TouchID which would be very nice to have in a Mac if it offered all the advantages it does in IOS devices. At the moment it appears that this requires some customization of the processor to maintain the required security. This would be valuable enough for me right now to go with an Arm based "Mac" like machine. Apple has added other IP to their Arm processors (purchased or developed in house) that I could see Apple wanting Intel to incorporate into their processors, the voice processing module being one example. In the end their are factors in play that could pull Apple away from Intel, especially if Intel doesn't get its act together. I'm not all that certain it will happen because these things are always in flux. The point is there is a real danger (to Intel) that Apple could offer a laptop that doesn't have i86 inside.
I say laptop because that is where a real advantage for Apple may be had with an Arm based solution. People need to remember that Apples Arm does very well in SOME bench marks against Intels offerings while running at a fixed clock rate. Intel on the other hand is running with a variable clock rate which at times runs far faster than Apples Arm. Of course one bench mark doesn't tell the whole story and frankly I wouldn't expect Apple to use the same SoC that they use in the iPad in a laptop. However given a few improvements, the ability to run more cores at a higher clock rate being a big one, Apple can deliver a decent chip for the Mac Book Air just about anytime they want too now. They simply have to desire to do so. It really isn't a technical question as they certainly have the engineering skills.