Slashdot Mirror
Opinion: Apple Should Have Gone With Intel Instead of TSMC
itwbennett writes "Apple is planning to have its ARM processors manufactured by TSMC — a move that blogger Andy Patrizio thinks is a colossal mistake. Not only is TSMC already over-extended and having trouble making deadlines. But Intel was clearly the better choice: 'Intel may be struggling in mobility with the Atom processors, but Intel does yields and manufacturing process migration better than anyone,' says Patrizio. 'While TSMC wrestles with 28nm and looking to 20nm, Intel is at 22nm now and moving to 14nm for next year. This is important; the smaller the fabrication design, the less power used.'"
Remember when Intel took the MacBook air design and turned it into the Ultrabook reference design for its Wintel PC OEMs? Why would Apple not want that to happen again, only faster?
Help stamp out iliturcy.
Sounds like a silly premise. Who says Intel would even want to do it? Why would Intel want to go back into ARM fabrication when they are trying to beat ARM chips with Atom?
People are buying the platform, and it only comes from one vendor. It's not like with Android where you can compare different hardware specs. Apple will produce a single product at a given price point with a given set of hardware specs, and that's what people will buy. Not saying this is a good or bad thing, just that it's a thing.
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
Intel supplies most of Apple's CPUs, yes?
To give one supplier most or all of your business gives them a HUGE advantage over you.
Just look at what happened to everyone who tied their business to Microsoft or IBM.
This is a business strategy issue - not a tech one.
Personally, I think Apple should take their cash and make their own processors, allowing for their OS to have a firmware component and thereby boosting performance and security.
They make their own chips, and you buy what they make.
Apple isn't going to be able to get Intel to fab their custom chips for them. That isn't Intel's business model.
Intel sells their own CPUs. They don't sell your CPUs.
They just happen to have the best fabs.
Hit submit too soon... replace "greatest" with "best"... I forgot my Apple lingo for a moment.
Help Brendan pay off his student loans
Intel's high end fabs are tasked to capacity with their own chips near as I know. They are probably not interested in taking on outside orders for ARM chips.
Now I suppose Apple could switch over to x86, but I doubt they'd be willing to do that given that they own a big stake in ARM. Also at this point Intel doesn't have x86 processors suitable for phones. They may make such a thing in the future but they do not now.
So ya, Intel would be the best option... if they were an option. They have fabs above and beyond anyone else, they spend billions in R&D on it and as such are nearly always a node ahead and have good yields. However, their fabs are for them. Their 22nm fabs are busily cranking out Haswell and Ivy Bridge chips. They are not for rent for cranking out ARM chips, unless something has changed since last I looked.
does Intel even fab for third parties? of course it would have been apple's first phone call if they did, even before using Samsung.
The post makes no mention of what the negotiations or considerations Apple or intel had - just that they spoke, at some point, about something. Intel walked away from their own ARM licence in the past and has not been known to fab many outside designs. Intel fabbing ARM chips for Apple would be quite a coup given their process advantage. Maybe even to the point of antitrust investigations if intel were not more inviting of manufacturing other competitor's cores.
Apple requires a huge number of wafer starts to accommodate their sales. Intel's existing 22nm volume, plus Apple's volume would exceed their built out capacity. Intel would undoubtedly require Apple to commit to a long term (eg 3-5 years) of 22nm volume purchases to offset their build out costs. Plus absolutely everything about Intel's flow is unique and different, you have to retool your entire design just for Intel. Samsung and TSMC are much closer, though still significantly different.
There are two interesting question sin all of this:
1) Where is the TSMC capacity for 20nm going to come from? Apple+Qualcomm alone is already a mind boggling level, let alone all the other 20nm customers TSMC has signed up..
2) Will Samsung have significant excess capacity at 20nm now? Will TSMC customers diversify toward Samsung?
We all know that Intel wasn't even an option. They're simply not in the business of fabricating third-party designs, for anyone.
What kind of ignoramuses write this shit? Do these writers get paid to write such nonsense?
Maybe I'm working in the wrong industry.
This is important; the smaller the fabrication.... the more static power used.
At 22nm, you hit v_sat before threshold anyway. The sweet spot for power is actually about 130nm.
Going with Intel would have been too expensive and would have been a terrible conflict of interest. Intel would gain early access to all their designs and could use it against them with their atom designs. They were avoiding the same situation they were previously in with Samsung. At first Samsung didn't really compete with Apple but things radically changed and using them as a foundry gave Samsung early access to Apple designs. Intel might not really compete with Apple now in mobile but that is surely going to change as Intel improves their low power chips. This also gives Apple the flexibility to consider migrating more platforms to ARM. So sticking with a company that doesn't venture outside of the foundry business is a safe business decision and Apple can use their leverage to speed up the R&D to make them more competitive with Intel (TSMC recently decided to accelerate migrations to 20nm and 16nm, I wonder who pushed that...).
Is that apple is saving money per chip by going with TSMC. I don't know that its true, but companies always have a strong motivation to attack the bottom line rather than go for quality. Steve Jobs was exceptional because he wanted quality in his products even if they ended up costing more. The exceptional part being that he was good at resisting the lure of cheaper components.
who prays for Satan? Who in 18 centuries has had the humanity to pray for the 1 sinner that needed it most? ~Mark Twain
From the article: "While TSMC wrestles with 28nm and looking to 20nm, Intel is at 22nm now and moving to 14nm for next year. "
TSMC's 28nm process is, in fact, widely considered a big success. Although it didn't ramp up initially, quite as a fast as their customers wanted, that only lasted a few months at start of 2012. Look a bit closer you see changing nodes has problems for all manufactures (even Intel).
20nm is in fact ahead of schedule. The likes of Altera are going to have to wait 2 years before they start producing chips on Intel's 16 nm process. While Apple will have 20nm early next year.
Apple could buy or merge with Intel, and then announce "x86, end of life, ten years. Merry Christmas, AMD." That would be the end of Microsoft, since nobody ever wanted Windows on anything other than x86.
Meanwhile, if Apple used Intel's fab for all of their processors, they could reduce their power consumption at a much faster pace than they're already doing. I'd love to get 20 hours of operation per charge from an iPad.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
"Apple is planning to have its ARM processors manufactured by TSMC — a move that blogger Andy Patrizio thinks is a colossal mistake.
Why would Intel want to manufacture ARM processors? They might make some money in the short term but the real profits are in owning the intellectual property behind the design. Intel would basically be subsidizing their biggest competitor. It would be akin to asking Microsoft to start their own linux distro or like Apple switching to Android. It makes their product undifferentiated and kills their margins.
Intel always has the option to start making ARM processors in the future but they'd be pretty foolish to do it at this point.
Intel gets high margins on much the x86 line. What on earth makes this douche assume that Intel would be willing to accept Samsung's margins in order to enable Apple to shift even more of the consumer market away from x86???
While you are correct that there are some people who will buy it regardless of what it built with, those days are fading, as many people are fed up with the slow pace of change in the Apple phone arena, and Apple wouldn't want to incur the delay penalty of a switch, when they can accomplish the same goals with their current hardware.
Many pundits are fed up with the "slow" pace of change in the Apple phone arena, because they need new clickbait twice a day. I have yet to hear from anyone who actually owns an iPhone that they're anything of the kind. Most people buy a new phone every 2 years or less frequently. The current iPhone is a significant improvement over the 2-year-old iPhone I have now.
Or are you trying to say that "many people" know or care about things like NFC, fingerprint scanners, or other check-boxable features that most people who don't read Slashdot have never heard of and wouldn't care about if they did? Because the actual numbers of iPhone sales don't seem to bear out that kind of view...
Dan Aris
Fun. Free. Online. RPG. BattleMaster.
Oh they don't. Then why would they manufacture for Apple? Intel's real edge is not processor design, it's manufacturing know-how. Watch PBS's Silicon Valley and understand that Intel is less about designing great processors and more about beating the competition with better processor fabrication. Intel is at least 12 months ahead of everyone and so why would they give that lead up for Apple.
I love the summary.. some random dude (blogger) that writes stories on the Internet has an opinion and thinks mega-billion company is making a mistake.. and this is news on Slashdot nowadays.
How low Slashdot has fallen..
Intel *is* a foundry. They make chips for third parties. They have a whole "Intel Custom Foundry" division dedicated to this. They make chips for Cisco, Netronome, Altera, etc. Some of those chips even have ARM processors.
Intel is inching into the foundry business.
They are *not* making chips for Altera. They have a deal with Altera to make chips at 14nm but Intel doesn't even have a production 14nm process yet. The Cisco deal was only signed in January. No word on when they expect to ship. Their shipping customers (Achronix, Tabula, Netronome) are all startups with limited volumes. Apple needs huge volume. I don't think Intel is ready for that yet.
Intel *does* make custom chips for outside people, contrary to what some people are saying. They sub out spare capacity, especially in older fabs. They just don't make them on their newest foundry processes (the ones that would be actually useful to a company like Apple) for a variety of reasons, the chief one being the newest processes are generally full to capacity. Even if there were some space available it wouldn't be near enough to satisfy Apple's demand for A-series chips. You have to remember, an A-series chip requires on the order of 150,000,000 units in the first year.
Apple has the cash hoard to get into the foundry business if they wanted, but it would take at least a decade to hire engineers and gain the experience necessary to cost-effectively produce stuff like 14nm 3D transistor chips, assuming you can navigate the patent minefield.
The only way such a deal would work is if Apple funded a new Intel foundry to produce Apple chips in some kind of long-term deal, but that would probably require Apple to spend double on the processors to give Intel the fat margins they want to even consider the idea.
I fully agree that in some magical world where this kind of deal happens it would give Apple a permanent advantage in the phone space, as no one would be able to come close to the performance and performance-per-watt of an A7-type chip made on Intel's latest 22, 14, or 10nm process; it would make all other phones look like a joke.
It just isn't going to happen.
Natural != (nontoxic || beneficial)
the smaller the fabrication design, the less power used
Ummm, no. The smaller the design the more leakage current you get and the more power is wasted as heat. Who is this idiot that wrote this completely clueless "opinion"? Intel does have a foundry unit, but they don't make lots the size that Apple would need with the fabs that Apple would want to use. The third parties aren't getting lots on 10 million finished parts per quarter at 22nm from Intel. Not unless those third parties are buying Intel branded parts.
I want workstation class ARM processors back. 16 core 4 processor behomith Motherboards to give us on the desk the performance we should have had a decade ago.
Do not look at laser with remaining good eye.
ITW bennett
Fuck me slashdot's editors are lazy.
I'm a good cook. I'm a fantastic eater. - Steven Brust
Yes, I'm sure that the (more or less) biggest company in the world, currently being run by the operations guy who helped them reach record-setting levels of profit in the last decade, did not do their homework when evaluating manufacturing partners. Thanks, random blogger guy, I'm sure they'll straighten all their shit out post-haste!
Or maybe, just maybe, the guy who runs one of the most successful companies on the planet and earns more in a year than you and your family could earn in ten lifetimes, actually knows what the fuck he is doing. Just a thought.
Dear Slashdot: next time you want to mess with the site, add a rich-text editor for comments.
The processor industry is full of players who spend billions of dollars trying to make marginal gains in performance. Spending tons of money on R&D for new processors makes sense if you're selling those processors to lots of other mobile electronics manufacturers, but it doesn't make sense when you're simply hoarding them for your own products. It especially doesn't make sense considering nobody is buying iPhones because of their processor specs, let alone the fact that Apple produces their own chips. Apple would be better off using commodity hardware and spending their money on improving other areas of the user experience. But since I have turned against Apple over the past few years, I'm just fine with watching them make costly mistakes. From what I've been reading, quality control at TSMC has been somewhat questionable and Apple is asking a lot from them to make these new chips with cutting-edge fab processes at a high volume and with minimal defects. I have a feeling this decision will cause trouble for Apple in the future and all of this is a result of Apple trying to punish Samsung in their childish feud.
Ultrabooks have been around since the 90's. only thing that changed is that intel is now making decent ultra low voltage CPU's and they use flash memory instead of HDD.
You're talking about subnotebooks, which were stripped-down ultraportables designed to save weight at any cost (including tiny screens seriously compromised performance.) They eventually evolved into Netbooks (remember those?) when the price came down.
"Ultrabook is a specification and brand developed by Intel for a class of high-end subnotebooks which are designed to feature reduced bulk without compromising performance and battery life."
Ultrabooks are not subnotebooks; they're designed to be as fast, if not moreso, than regular notebooks. Unlike subnotebooks, they have fairly fast CPUs and their screens are as large as entry regular notebooks (13") with even higher resolutions.
Basically, they all have lightweight designs ( 4 lbs) and no optical drive
Subnotebook:
High price, tiny screen, slow CPU, good build quality, good for typing and not much else
Netbook:
Low price, tiny screen, slow CPU, shitty quality, good for web browsing and not much else
Ultrabook:
High price, midsize screen, good CPU, high build quality with unibody case, good for pretty much everything except gaming
In typical Slashdot revisionist fashion, when Apple introduced the MacBook Air, it was a strange, dumb idea, because only an idiot would pay top-dollar for a machine with no optical drive and no upgradablility, and nobody but Apple would be stupid enough to try such a thin. Now, of course, it was an obvious development that was actually no different from what everyone else was doing.
Many people are posting as if Intel would be involved in chip design (for example: "Intel doesn't want to make ARMs."). Intel would be acting only as a foundry: Apple does the design work, sends Intel a set of files specifying mask geometry; Intel makes masks and fabricates the chips.
The questions thus become, who has good enough technology and who is a reliable supplier? If Apple doesn't need the finest tech that only Intel can provide, then using Intel isn't necessary.
TSMC having production capacity limits can be a problem, and it's likely to have delayed deliveries in a crunch. But foundry is their only business, not producing is not an option. Intel on the other hand, can decide "we need all foundries for internal use. Make your lifetime orders now; no new business will be accepted." 33 years ago (the only information I have, from a then-Reticon employee) was that this was a substantial risk in dealing with Intel as a fab.
Contribute to civilization: ari.aynrand.org/donate
There is a cretinous technology meme that states that Intel is the industry leader in fabrication technology. Nothing could be further from the truth. Intel's fabs are good for one thing, and one thing only- Intel's CPUs. Intel optimises every aspect of its fab technology for a single type of microprocessor, and does not care that its fabs are thus a horrible match for any other type of integrated circuit.
Intel's strategy, of course, is based on the fantastic profits that Intel earns from being a de facto monopoly producer of x86 chips. Yes, technically AMD makes x86 parts as well, but their share of the market is tiny, and a constant tiny at that. Intel cares only about the x86 parts, and their ASP (average selling price).
The downside for Intel is that they have ZERO flexibility if the marketplace suddenly changes. Their fabs are horribly expensive, driven by massive R+D investments paid for out of the phenomenal profits from their x86 chips. Intel has not needed fabs operating at commercial rates for decades. They simply don't compete in that way.
TSMC is the very opposite. TSMC has incredible experience making the highest performing GPU parts. TSMC has worked with the greatest number of diverse chip designing partners. And TSMC, having badly slipped with its 20nm process, is currently losing a lot of its customers to Global Foundries.
Apple is effectively buying TSMC. It isn't an actual purchase in the traditional sense, because of how Taiwanese businesses operate, but to all intents and purposes Apple has decided it wants its own fab company. Obviously, in no sense was Intel even in the running. Intel is a bad joke.
You see the same thing with the new consoles from Microsoft and Sony. None of the tech or fabrication runs comes from Intel. Intel lacks any technological expertise to serve this market. It can't design the chips. It can't produce the software (low level) for the chips, and it certainly cannot manufacture the chips.
None of this matters unfortunately for Intel. Intel is in an absolute disastrous period with respect to its fab technology. Intel bet the farm on FinFET, and both its attempt to build FinFET based x86 parts have been total failures. Sure, the chips 'work' but they are slow (maximum obtainable commercial frequency), hot, and do not give the improvement at low energy usage Intel was banking on in the first place. And Haswell was Intel's second FinFET design, which proved to be even worse than the first attempt at FinFET with Sandybridge.
In fairness, the entire industry thought FinFET to be the future- it is just that Intel had the money to try to be first. Now FD-SOI is proving to be the wonder technology on current geometries- and at a fraction of the 'cost' of FinFET. TSMC and Global Foundries are both far ahead of Intel in this tech.
Intel is in such disarray, it has cancelled (for the first time in forever) its yearly release of new desktop parts in 2014. Intel is desperately trying to decide how to build its new parts on the next shrink, and also seems to be in (not so) secret partnership with Nvidia to use Nvidia's graphics in parts released 2015 or later.
Apple wants to be the new Qualcomm, not the new Intel. Apple wants to go fully ARM as soon as feasible, and this move would seem to coincide with Apple's move to ARM's V8 architecture- the 64-bit version of ARM that matches x86 in memory addressing capability. Now who in their right mind would attempt to build first-quality ARM SoC parts on an Intel fab?
They have an ARM license and I think they tested a 28nm SOI ARM design a while back. They may not be as advanced as Intel, but I don't think they are over capacity either.
You should think that through a little bit more. It would take less than 10 years for MS to covert all their apps to non-x86 platforms (the kernel and Office already run on RT machines). If x86 dies Microsoft can drop all legacy support, which is the source of most of MS' problems (excluding Metro) and gets to resell a new license for every piece of software to everyone. MS will come out with a stronger product line and more cash. Any companies complaining about the switch and lack of backward support will be pointed towards Intel/Apple for killing x86. Microsoft gets off free and Intel/Apple become the bad guys in corporations' eyes.
This is business/fanboy news, not tech or science news. Fuck off, Andy Patrizio.
Merry Christmas AMD indeed, provided they could survive the ten years needed to become the dominant x86 manufacturer. Meanwhile, you could expect x86 front ends to pop up on fundamentally different hardware, even if it's inefficient. This move might increase the adoption of non-x86 platforms in the office, and get Microsoft to write for them as well (having a fully operable Office would be enough for a lot of businesses), but AMD would be stronger than they are now.
How is the Riemann zeta function like Trump rallies? Both have an endless number of trivial zeros.
It would take less than 10 years for MS to covert all their apps to non-x86 platforms
Been there, done that, failed miserably. Remember Windows NT for SPARC, Alpha, etc, etc? Remember all the people not buying it?
MS is all about the legacy.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Intel abominates ARM. They will do anything for it to die. Intel's marketing dept will not let them fab any ARM. Thats why they sold their ARM division.
I am sure that Intel offered Apple to use their x86 platform on the iPhone and iPad, but for some reason (energy efficiency, price, etc) they stuck with ARM.
Also it seems Apple is doing really well with ARM business, so I don't think they plan to make the switch anytime soon.
while funny think about this...
I know lets buy out one of our suppliers then shut them down so there is less suppliers we can play against each other. /s
ARM can be pretty much made at any foundry. x86? Not so much. The only reason they moved to x86 in the first place as powerpc was getting embarrassing with heat and speed. If ARM was decent on the desktop they probably would move to it.
TMSC is a Chinese company in Taiwan.
Why become yet more dependent on China?
Andy Grove once espoused paranoia. Now Intel espouses dependence.
Short-sighted.
TSMC (Taiwan Semiconductor Manufacturing Corporation) is a Taiwanese company, not Chinese. Unless you are going by Beijing's claim that Taiwan is a part of China, but even then, the first statement doesn't make sense.
If people had no need for Windows on SPARC, Alpha they would not and did not buy it if Apple shut down x86 production in the scenario above people would have to switch chips and would buy Windows on a new platform.
The fact of the matter is Intel has no foundry presence or experience. NOTHING like TSMC who is the #1 volume and revenue leader in foundry services. The technology aspect is transitory if Intel has an "advantage" at all - TSMC and Intel have different strategies and of the two, TSMC's is actually more conservative. So you have a company with lots of captive experience and NO foundry experience, a business model that depends MORE on logistics and relationships than it does the technology. You CAN NOT just "Become a foundry tomorrow because we decided it was a good idea".
The answer to this is pretty obvious: Apple called it right.
BTW I have >30 years experience in semiconductor design and management, both in captive and foundry work.
Apple has two major 20nm/14nm FinFET Fab giants to choose from without taking it in the rear from Intel, while dictating their designs. Sorry, but Intel will never get Apple's embedded space manufacturing.
Arguably this honour goes to ARM
Uhm, try again...
OK, I'll try again...
ARM has shipped over 30 billion processors with a current annual rate of 5 billon per year. Only 3.5 billion PCs have ever been sold and the current rate is less than 0.5 billon per year. So, yes, I believe it is arguable that the ARM ISA has sold more than x86.
FTA: "This is important; the smaller the fabrication design, the less power used."
who are you lecturing? goofy $%$^%
NT was never publicly available on the SPARC. Intergraph toyed w/ the idea some, but then dropped it, and went whole Wintel. NT debuted with the MIPS and Alpha, and added PPC later. And then dropped MIPS, PPC and Alpha, in that order.