Categories

In search of Apple's microprocessor

Digitimes Research published an estimate of the allocation of production of application processors by Samsung. It suggests that Apple will transition away from Samsung as a processor supplier during 2013 and stop sourcing altogether by 2014.

Most current capacity (nearly 70%) is said to be allocated to Apple’s products. The iPhone, iPad and iPod touch consumed an estimated 226 million processors in 2012. That demand is expected to increase but it will increasingly to be met by other suppliers in 2013. However, Samsung’s own products will take up some of the slack.

The following graph shows the estimated production schedule by customer.

Screen Shot 2012-12-28 at 12-28-9.57.37 AM

There are several questions that this data raises:

  • Does Samsung’s microprocessor production exceed that of Intel?
  • Who will produce Apple’s processors after 2013. Will it be one or more suppliers is also an open question.
  • Is Apple’s CapEx (est. $10 billion/yr.) being spent on some of this future capacity? A modern fab costs about that much.
  • As the iOS processors are Apple’s own designs, will Apple be integrating production and design?

Prior to his departure Intel’s Paul Otellini had begun speaking about manufacturing. He argued that the “fabless” model was no longer competitive and that design and manufacturing should be re-integrated.

I think we’re about to find out to what degree that will happen.

 

  • obarthelemy

    I’m not sure we should give much credit to the pronouncement by the CEO of the major integrated chip manufacturer that “integrated is the way to go”.

    Apple certainly have the volume to justify a fab, but chip manufacturing is not a high-margin business, and there’s several suppliers already. Unless Apple think they can do it better, or identify that as a strategic part of their supply chain, I see no reason why they would want to lower their margins and get distracted by getting into that business ?

    • Luis Alejandro Masanti

      To maintain chip design secrets —specially in Apple’s case— could be another good reason to own the factory.
      Horace had good points (in other posts) about the learning by those in the lower end of the chain,

      • http://twitter.com/fivetonsflax fivetonsflax

        And perhaps the order size and timing as well.

      • obarthelemy

        I’m really not sure it’s that meaningful
        1- all leading chip suppliers are pretty much at parity, performance/features/power consumption-wise. Apple do not have a significant advantage in that area, let alone a performance leadership (which is not what they’re aiming for, anyway)
        2- assuming there’s reverse-engineering going on, by the time it’s done and can be integrated in new product design, 1 or 2 generations of chips have gone by and it’s mainly irrelevant
        3- Using Samsung as a fab was an especially funny move. Other fabs are independent of any competitor (TSMC, UMC, Global Foundries…)

        I’d think screen technology would be more important to Apple. Retina got more buzz than A6X.

    • http://www.noisetech-software.com/Home.html Steven Noyes

      100% agree on this. I would see this as a distraction for Apple.

    • Walt French

      It’s not too hard to fantasize about future product & service offerings that might have Apple wanting many, many millions of custom CPUs, with little or no warning to competitors, and little or no leakage of concepts such as a completely different instruction mix or optimizations.

      Siri- or at least full-voice-recognition-in-the-phone, anyone? Dedicated HTML- or VM-processing instructions? Clusters of edge processors at each cell tower to speed data transfer, e.g., instant webpage views? A CPU with no graphics but audio-style signal processing for a screenless smartphone? New audio codecs that’d allow carriers with Apple boxes at their data centers to offer higher audio quality?

      Apple is pretty much astride the tiger of innovation today—get off and it’ll get eaten by the hungry smartphone industry that can do just about anything with a couple of years’ notice. Having the CPU capability inhouse gives them the option to make dramatic changes in that aspect of their product.

      • Tatil_S

        Fabs do not need to know the function of your chip if you don’t want to tell them. Whether you use your brand new chip for a handset or at a cell tower or in a TV is your business.

  • Luis Alejandro Masanti

    As for the question on Samsung vs. Intel chip production, are cellphone/tablet chips the same/different size than Intel desktop/servers chips?
    At the end, I think, what matters (from this point of view) is the total amount of chip surface that counts.

  • http://sumocat.blogspot.com Sumocat

    Good data as usual, but why would Samsung cut their own chip production rather than cut off their suppliers? Per the Digitimes report, “Samsung’s in-house capacity for APs currently fulfills only about 30% of the total demand for the firm’s own-brand smartphones and tablets,” primarily their high-end Galaxy line. The remaining 70% (same as Samsung chip capacity allocated to Apple) is produced by outside vendors. Samsung’s products could pick up all of the slack if Apple left, not just some of it. Perhaps that would not be the most economic option, but Samsung’s obsession with in-house production cannot be overstated.

    • Canucker

      How do you reconcile 70% of Samsung’s chip production being produced by outside vendors with Samsungs “obsession” with in-house production? Perhaps they were saving money by out-sourcing? Where is the evidence that the capacity released by Apple’s withdrawal is appropriate for the out-sourced chips? Samsung clearly will sell chips to anyone so I’d guess some of the newly released capacity will go towards chips for other Android OEMs (who are actually more direct competitors to Samsung than Apple). Samsung is certainly a very interesting company – very different from Apple, etc. The next few years will be interesting.

      • http://sumocat.blogspot.com Sumocat

        “How do you reconcile 70% of Samsung’s chip production being produced by outside vendors with Samsungs “obsession” with in-house production? Perhaps they were saving money by out-sourcing?” — Almost certainly it is more profitable to sell high-end chips to Apple and buy low-end chips from outsiders. As explained in ‘The real threat that Samsung poses to Apple’, Apple is essentially funding Samsung’s rise in the mobile market. http://www.asymco.com/2012/12/07/the-real-threat-that-samsung-poses-to-apple/
        But that cycle ends if Apple stops buying chips. Then it becomes a choice between buying outside chips or letting their own chip production go underused. For other companies, that would be a simple analysis. But Samsung produces food to feed factory workers, equipment to build factories, equipment that goes into the factories, trucks to take products from those factories to ships that were also built by Samsung. Samsung does all that and more. That is why I call their in-sourcing “obsessive”. One underused Samsung factory affects far more than any one division of the company.

        “Where is the evidence that the capacity released by Apple’s withdrawal is appropriate for the out-sourced chips?” — Today’s high-end chips are tomorrow’s low-end chips, meaning whatever high-end chips Samsung makes today will be considered low-end in 2014. That’s how Apple does it. Samsung might, if they were the least bit inclined to copy Apple, follow a similar path.

        “Samsung clearly will sell chips to anyone so I’d guess some of the newly released capacity will go towards chips for other Android OEMs (who are actually more direct competitors to Samsung than Apple).” — You mean the Android OEMs currently being crushed by Samsung? Samsung holds nearly as much Android market share as everyone else combined and shows no sign of reversing.
        http://techcrunch.com/2012/12/19/millennial-samsungs-android-market-share-grew-from-23-to-46-in-one-year/
        Unless there is a catastrophe, like the Koreas restarting the war, by 2014, the only Android OEM capable of taking meaningful advantage of Samsung’s chip-making capacity will be Samsung.

        “Samsung is certainly a very interesting company – very
        different from Apple, etc. The next few years will be interesting.” — Absolutely.

      • Tatil_S

        “Almost certainly it is more profitable to sell high-end chips to Apple and buy low-end chips from outsiders.”

        It might be, but that is not what Samsung is doing. It is buying plenty of high end processors from outside vendors. For example, North American Galaxy S3, its flagship phone, uses Snapdragon.

      • http://sumocat.blogspot.com Sumocat

        High-end is relative. The North American Galaxy S3 uses dual-core Snapdragon, but their other S3 variants use significantly faster quad-core Samsung Exynos chips.
        http://browser.primatelabs.com/android-benchmarks
        Nor does dual-core Snapdragon match the GPU performance of the Samsung-produced A5 (nor the faster 32nm version or the A5X).
        http://www.differencebetween.com/difference-between-apple-a5-and-vs-qualcomm-snapdragon-s3/
        Now, I’m not saying Snapdragon is low-end, but those particular chips do not match what Samsung produced for themselves and Apple, and the benchmark chart consistently shows that Samsung’s own sourced chips are better than what they outsourced.

      • Tatil_S

        Quad core chips may be “faster”, but they reportedly reduce battery life of the phone, so “better” is a bit of a stretch. This is just a trade-off. I don’t know how GPU performance of Exynos compares, so I cannot comment on it.

        Fabs get paid by the wafer. A low end, small size, but high volume chip can bring in as much revenue as a high end, large size, but low volume chip, as long as both use the same process node. A truly low end chip that uses an older process node is most likely not competing for capacity with products at higher end, as the older process nodes do not get retired for a long time and can co-exist with the new nodes for a decade. I don’t think the decision to outsource or not is based on trading high end with low end. (TI’s chips got into Galaxy Nexus, the flagship at the time.) It sounds more likely that based on release schedules and price negotiations, each development teams pick the best supplier. Not hobbling one division to make another division more successful is a smart strategy after all. Sony could have been in a much better situation now.

      • http://sumocat.blogspot.com Sumocat

        I apologize for my imprecise language, but we lack precise data on the metric in question, which is price. Thus, I chose to look at performance as price generally follows performance: higher performance = higher price. Furthermore, the chips Samsung produces for Apple are custom designed, whereas their purchases from other vendors are not, and custom designs typically cost more than stock. The chips Samsung buys should cost less than what they sell to Apple or to themselves.

      • Tatil_S

        Custom designed by Apple (reportedly), so Samsung should be getting paid for “cost of manufacturing the chips” and some fab margin on top, but when it is buying a chip it needs to pay for cost of design, as well as manufacturing, and a profit margin on top of both costs. Hence, it is quite unlikely Samsung as a whole will make more money by fabbing somebody else’s design and buying a chip with similar functionality from somebody else. The fab division, if sufficiently independent, would not care about the overall picture, only its own financial picture, so filling its capacity with A6 may make more sense than making Exynos. (Even if it is not as profitable in the short term, I believe it can be more healthy to run disparate divisions independently, so I don’t think it is a bad move.)

      • Carpenter

        Snapdragon is the only exception. They are probably using is mainly because on Qualcomm modem tech is very handy when competing in NA market.

        Then there is NovaThor from STE. They are using U8500 for few products like Galaxy Advance and Galaxy S3 mini. Mainly because of price but It’s also pretty competent single chip platform, I guess. It’s interesting that Samsung isn’t making single chip solution yet.

        Others like Broadcom chips are for low end and probably selected for for cheap price.

      • Tatil_S

        Galaxy Nexus, when it came out, was a flagship both for a new Android version, so Google and Samsung. It uses TI’s processor, hence S3 is not the only exception.

      • http://sumocat.blogspot.com Sumocat
      • Tatil_S

        I don’t think that matters all that much. For example, Galaxy S2 variants use all three processors.

    • http://twitter.com/WalterMilliken Walter Milliken

      I suspect the outsourced chips are probably lower-end processors using older process lines. Those are more forgiving of process problems, and possibly not worth using Samsung’s own lines, assuming those are mostly designed for smaller feature sizes. Lower-end processors also can be stamped out using cookie-cutter IP blocks, where the higher-end ones probably tend to need more tuning for both process and efficiency.

      • Tatil_S

        I once checked out which APs Samsung uses for its high end, well marketed products, Galaxy S2 or 3, Galaxy Nexus and one other, maybe Note, sorry it’s been a while. Each one used a different processor according to tech reviews. One was using Samsung’s own CPU, one was using Snapdragon and another was using TI’s.

    • theothergeoff

      One thought is that outsourced capacity is often multi-year, and until Apple officially gives notice they can’t stop extending those contracts (Samsung builds lines for Carriers for several years, and likely have planned 3 year production runs, cast in an initial contract, and 1-2 contract extensions (a 1 year initial run, then an extension during that run that sets the ‘production run rate’ based on actual sales, and one more if there are long term supply needs). I can’t see a massive multi-line manufacturer being able to reel in capacity any faster than it’s current contract expirations, or face early cancellation fees.

      • http://sumocat.blogspot.com Sumocat

        Agreed, but that should apply to both Apple and Samsung. If Samsung could lose Apple in 2013, they would know that and could have timed their contracts with their chip suppliers to end concurrently. I see no reason to believe Samsung would be caught with their pants down and let more than half their chip production capacity go to waste if Apple decided to cut ties.

      • theothergeoff

        I don’t think big business works that way, especially interdivisionally.
        The chip division can’t force current contracts to just break off (assume a 365 days of contracts being signed.. each with a 1-3 year commitment [think options to extend*])… They have to honor the entire contract, and the option may be too good to refuse from a profit margin perspective of the device division (whose bonuses are tied to their division performance, not the chip division).

        And I can’t see Samsung, sitting with an Apple option to extend in hand (which my understanding is their modus operandi), can in good faith overbook all that internal capacity until the option expires.

      • http://sumocat.blogspot.com Sumocat

        Yeah, but how much honor and good faith should we really expect from a company run by a twice-convicted (and twice-pardoned) criminal?
        http://www.engadget.com/2009/12/29/samsungs-former-chairman-pardoned-again/
        As Canucker put it, Samsung is a very interesting company. Definitely not like others.

  • Walt French

    Disruption Theoryspeaks to why integration is, or is not good, doesn’t it? Maybe the answer relates to how quickly Silicon technology is changing and the benefit of keeping it integrated to design.

    Otellini would be right if there are significant new advances coming. But if there are only incremental changes coming in Silicon design, the fabless design will continue to succeed I think.

    • Canucker

      Depends somewhat on the pace of iteration. Otellini was, of course, defending Intel’s strategy against their nightmare – being reduced to a glorified, high tech stamping press with no control over its destiny. The problem with owning a FAB is the temptation to prolong design cycles in order to maximize profit from a design (given that yields typically increase and costs decline). Companies like Apple seem to be happy with a new chip design per device, which looks like its moving to a 6 month introduction cycle (with 12-36 month production). They shop around. This is largely driven by the fact that we’re still in an exploding performance market where improvements are doubling. This significant refinement may go on for another 5 years – especially in matching energy reductions. At some point, the pace will slow as the components mature, but not yet. Its a buyers market.

      • Walt French

        The problem you identify of fab owners being tempted to stretch out the design cycle is real. But I’d think that an integrated design/process/customer like Apple (or Samsung without Apple) has a strong incentive to maximize OVERALL profitability, and that the relatively small impact on a phone’s overall margin would bias them towards more flexible designs as they discover/invent new applications or services.

        I couldn’t* edit my post this morning, so I wasn’t able to note that much of this issue is outlined by Coase (a Nobel laureate—I love groundbreaking work) in his Nature of the Firm. Perhaps that was adopted wholesale into Disruption Theory, which tackles another aspect of the Theory of the Firm. But I think that inasmuch as CPUs are a small fraction of a smartphone’s current price and a modest share of its energy budget (the screen being the big drain), and faster CPUs end up making a large impact on user perceptions, we’ll reach the “good enough” point and maybe blow through it, fairly soon.

        Maybe you could clarify what you mean by it being a “buyers’ market” today. Handset makers are scrambling to differentiate and/or improve their products, with market share, even viability on the line; striking a deal to get a next-gen design months ahead of others is a Big Deal. That’d seem to put Samsung, Qualcomm and others into the driver’s seat, while the Intel threat is not so strong that they’re lining up customers.

        * Pro Tip: don’t dictate Disqus posts from the browser built into your iPhone’s Bloomberg Anywhere News Monitor, while you’re in a noisy coffee shop.)

      • Canucker

        Buyers market insofar as there are alternatives and significant over capacity in FABs. Also, as we’ve seen, Apple seems to create a perennial buyers market for itself insofar as the company is willing to invest in off-setting the up-front capital costs *and* guarantees a large order (basically defined as as much as you can feasibly produce). It’s different for the second/third tier customers who impatiently pace around the carcasses waiting for the lions to tire of eating and then plod off to snooze in the shade.

      • Tatil_S

        Over capacity? There is a big crunch in 28nm fab capacity that is only now coming to some balance. If you want to be at the forefront of process technology, there is not a whole lot of places you can shop at.

    • http://twitter.com/WalterMilliken Walter Milliken

      I think both sides may be right on this. Fabless design was partly driven by the fact that there were far more people capable of designing silicon than could afford to actually build process lines. And not everyone made good use of their process lines’ capabilities. Also, I think the integration only helps with the very cutting edge designs, where tweaking both the design and the process can buy you benefits.

      But since the cutting edge is constantly moving, there’s a big capital cost to staying there to make use of the integration synergies. It seems that only Intel with their high-priced CPUs could make use of it.

      In my opinion, there will come a point fairly soon, though, that it’s going to be hard to use chip area (at least on CPU chips) for anything that benefits performance much more. Right now, Intel’s strategy is putting more cores on a chip. I expect this to top out at about 8-16 cores for architectural reasons, except in trivially-parallel applications that are primarily compute-bound (as opposed to memory-bound).

      Once more cores don’t add anything, about the only thing they can use more silicon area on CPU chips for is cache memory. Which is useful, to a point, but may run into diminishing returns fairly quickly.

      Frankly, I think the place where a integrated design may really be beneficial is redesigning the signal interconnect and packaging. A lot of CPU power now goes to driving interface pins. Keeping all the pieces on-chip both speeds things up a lot and lowers power consumption. But you have to be building everything as an SoC at that point, which favors customized designs like Apple’s A6. You still don’t need to own your fab, but it does make it easier, since you have to re-do the synthesis and masks for each process you use.

    • http://twitter.com/handleym99 Maynard Handley

      I can’t speak to the manufacturing side, but on the CPU micro-architecture side, there are still a number of interesting ideas for improving CPUs that have never been implemented.

      The problem has been that Intel and IBM are both constrained along particular paths.

      IBM (with POWER4,5,6,7,8) sells a very expensive CPU in a very expensive box to people who can’t do what they want with a cheap box. That means IBM is primarily interested in making its CPU able to handle ever larger quantities of bandwidth (which is the expensive part that you can’t get in a cheap box).

      Intel has two big constraints. One is that their designs are so complicated (it’s now 7 years+ from start of design to end) that they proceed largely on guess-work — the turn-around cycle between learning that some micro-architectural idea is not very good (or IS very good and should be used more widely) is horribly long. The second is that they really want their designs to scale all the way from a low-end laptop to a high-end Xeon. They do what they can to handle this via multiple cores, adding extra memory controllers to the Xeon, stripping SMT from the lowest i3’s, etc; but they can’t have radically different low-level guts for these different targets.

      Apple does not suffer from these various constraints. They can design to a very particular target, and their designs are not (yet) so complex that they take many many years to spin. Which suggests that there is value to their custom-designing what they need.
      HOWEVER a design without an awesome fab capability is of limited value. So everything I’ve said is moot in the absence of such fab capability…

  • gprovida

    An interesting consequence of this is the investment Apple is making in SAMSUNG competitors and these competitors will be competing on more than Apple products. In addition, Apple is the kind of customer that allows a component manufacturer to innovate new products and technologies with smaller risk, that is, a great customer, who would probably co-invest for early and exclusive access to the risker and more costly components. This is a much bigger loss [partnership with Apple] and likely stronger and more aggressive competitors who will service much more than Apple. That is, take business from SAMSUNG on components as well as strengthen end product producers e.g., HTC and LG.

  • http://www.facebook.com/profile.php?id=741717344 Dick Applebaum

    I think several things are in play here:

    1) Apple’s overall need to increase manufacturing capacity for emerging [smartphone] markets like China and India.

    2) Apple’s desire to remove Samsung from the critical path of their supply-chain… For obvious reasons.

    3) Apple’s apparent move to a two-release-per-year iDevice product cycle — with two ARM processor/package upgrades per year.

    4) Apple’s apparent move to offering an increased number of older iDevice models to cover more price points — providing less of a price umbrella for the competition

    5) Apple’s desire to bring some of the manufacturing to the United States — And, take a bigger part in that manufacturing process.

    6) Apple’s dedication to the continuous expansion of the iDevice/iCloud ecosystem to make it a major consideration of the buying decision for Apple devices.

    The way I see this playing out is that Apple will continue to build older models of iDevices on existing manufacturing lines — Mainly Sammy, for A6, A5 and A4 processors.

    As new processors are introduced, say the A7 and A7X in early 2013, then the A8 and A8X in late 2013… As much production as possible will be done by other manufacturers instead of Sammy.

    Rinse and repeat… After a few years Apple will totally be weaned from depending on Sammy.

    • oases

      What does No. 6 have to do with processors?

      • http://policydiary.com/ John S. Wilson

        What he said here: “The way I see this playing out is that Apple will continue to build older models of iDevices on existing manufacturing lines — Mainly Sammy, for A6, A5 and A4 processors” That means that more capacity will be necessary as older processors are used.

      • http://www.facebook.com/profile.php?id=741717344 Dick Applebaum

        I should have fleshed that out a little more… I think that Apple may flesh out its offering of older iPhone models. The current [older] iPhones are offered with a single storage choice — 16 GB for the 4S and 8 GB for the 4.

        By adding lower-price storage bumps, say $50-$75, for older technology they would have offerings at more (and lower) price points — especially for markets not currently penetrated by the iPhone… India, etc.

        Along with this, Apple continues to expand iDevice/iCloud ecosystem — free iCloud accounts, more services such as movies to more countries, etc. This makes all models of iDevices (even older models) more attractive than competitive devices. Though, iCloud to some extent, mitigates the need for more storage…

        Competitors release multiple phone models throughout the year — and currently being sold new.

        Apple will compete with these — using older models with the latest iOS — better hardware, better OS, better ecosystem.

        I think Apple’s ecosystem is part of the overall strategy that will allow Apple to wean itself from Sammy [producing Apple processors] over time.

      • tmay

        No seer, but I would have to imagine that Apple would have a hardware team looking at SoC’s as application processors within Lightning docks or Thunderbolt stream processors for Apple’s many Pro and consumer applications.

        As an example, an iPhone/iPod/iPad Lightning dock with an SDXC slot might include an A6X SoC as an application preprocessor for iMovie, iPhoto, and Garageband or provide transcoding for files to Pro Res for upload to FCX or RAW processing for (the rumored) Aperture X. With DSLR’s soon to exceed 40MP 16bit images, iDevices would have more “X” for processing without the penalty of battery consumption or increased weight and volume.

        Similarly, a Thunderbolt processor inline could provide some added processing performance no different than Kepler based GPGPU’s, but perhaps based on multicore 64 bit ARM architecture on a large die for entry level units. This would add value to the existing Mac lines and would repurpose older “A” series processor’s similar to Apple’s Atv.

        With Apple’s increasingly vertical stack of OS, Software, Hardware, Applications and media building out an even more desirable ecosystem, it places higher demands on competitors to compete as much on ecosystem as price.

        As to Intel, the lack of custom and proprietary X86 designs with very high premiums for is a major factor in Apple’s mac margins. I don’t see Apple exiting X86 anytime soon (arguably, Apple will follow MS’s lead if it comes to that).

      • http://www.facebook.com/profile.php?id=741717344 Dick Applebaum

        Those are some very interesting ideas…

        For an additional $25 parts cost you could have an intelligent/powerful lightning dock.

        For an additional $50 parts cost you could have an intelligent/powerful Thunderbolt IO connection.

        The Mac Pro scheduled for 2013 could use Ax ARM chips in addition to and/or instead of dedicated graphics cards — both inboard and outboard.

        A dedicated Apple Home Server, likely, could be entirely ARM and iOS based.

      • http://www.facebook.com/people/Shameer-Mulji/1685212657 Shameer Mulji

        “A dedicated Apple Home Server, likely, could be entirely ARM and iOS based.”

        This would be nice.

      • http://twitter.com/matthewwanderer matthew

        This Apple home server is otherwise known as the forthcoming Apple TV. IMHO.

      • http://www.facebook.com/people/Shameer-Mulji/1685212657 Shameer Mulji

        You quite possibly could be right.

      • http://twitter.com/WalterMilliken Walter Milliken

        I wouldn’t be surprised, but we may also see more auxiliary devices like the Airport Express and the Apple TV, with more of a distributed computing software approach in the “main” devices (Mac/iPhone/iPad). Essentially a more aggressive application of technologies like Grand Central Dispatch and LLVM compilation, delegating application functions to external devices (either bus-attached via Lightning/Thunderbolt, as you suggest, or via networking protocols like AirPlay).

  • http://twitter.com/WalterMilliken Walter Milliken

    I expect Apple will want to use at least two separate fab lines, for risk reasons. I assume Samsung as sole-source was only barely tolerable because they had geographic diversity of fab lines (Austin and Korea), and that may partly be why Samsung opened their Austin fab.

    If Apple does go to two suppliers, they will almost certainly have to make two different mask versions of the A6 (or A7), since you tend to have to synthesize and optimize to a specific process when you generate masks, at least if you’re pushing the limits of the process. They’ve done this in the past (the A5 die shrink), so it’s not impossible, but it does add a fair amount of engineering pain and cost. But I don’t think Apple wants to be at the mercy of a single fab supplier, especially after their issues with Samsung.

    • Tatil_S

      The effort involved in die shrink within the same process “type/fab” is not as big as having one state-of the-art processor being manufactured in two different non-compatible fabs. There are fabs that promise to match or emulate TSMC, some months behind of course, but I am not sure if Samsung and TSMC are all that similar. Intel is in a whole another boat.

      • http://twitter.com/WalterMilliken Walter Milliken

        One dual-fab case I know of was Xilinx, which was using both the TSMC and IBM foundries back a few years ago. I don’t know if they tried to run the same masks on both fabs, but I rather suspect not. And I don’t know how similar their processes were. Xilinx also had some major problems with their bleeding-edge FPGA features around the same time, which might have been related to the dual-vendor situation, or may have been something they just messed up on.

    • JimmyJ

      If I may – what is a “mask”?

      Thanks

      • Relayman5C

        The masks are used to define the components of the processor. If you use the same masks, you’re using the same design. Different masks, different design. So, in this case, you could replace “mask” with “design” and get the same meaning.

      • http://twitter.com/WalterMilliken Walter Milliken

        If by design you mean “low level transistor and interconnect layout”, yes.

        But most chip designs are created at the logic level in high level languages, and then compiled to silicon for a given process line (producing the masks). Then there’s usually some tweaking that has to be done after the fact, because the compilation process isn’t perfect.

        This is very similar to producing a piece of software for a computer — in theory, a program written in a high level language is portable across different machines and even OSes, but in practice there are a lot of issues. This is why moving to a different fab process is significant work.

      • sam

        The logic of a processor is common to different process/fabs, but after the logic is reduced to the graphic images (masks) that are used to expose the silicon, it would be different for each process level.

      • Walt French

        Masks are part of the very complex process of creating semiconductor chips. They control how different places on the chip are turned into insulators, etc. Best to get a bit more info than you think you want at Wikipedia’s Semiconductor device fabrication

      • JimmyJ

        Wow, way more complicated than I expected. I will indeed check this out on Wikipedia.

        Thanks Sam, Relayman5C and Walter Milliken for your helpful responses – folks like you (& Horace as well) are why I come to this site everyday.

  • http://www.noisetech-software.com/Home.html Steven Noyes

    I think a better question really is: Will Apple transition away from Samsung as a fab?

    Personally, I don’t put as much weight in this as many in the industry do. If Intel or TSMC or some other fab can take up that HUGE amount os production (and this is no mean feat) and do it at the same cost, Apple will transition. If not, Apple won’t.

    As for the question: “Does Samsung’s microprocessor production exceed that of Intel?” This is an easy question. From a numbers count, beating Intel is a simple feat to do. In 2002 desktop processors accounted for only 2% of the worlds supply of micro-processors with the other 98% headed toward embedded applications. I can only suspect these numbers are closer to 99%-1%. I don’t know yet if I call tablets/smartphones embedded or not. I sure treat them that way when I program but they blur the line in a serious way.

    BTW: What does “That demand is expected to increase but it will increasingly to be met by other suppliers in 2013″ mean? Do you mean: “That demand is expected to increase but it will [be] increasingly [difficult] to be met by other suppliers in 2013″

  • Pingback: Apple’s Chip Manufacturer Move Unlikely to Affect Samsung | HDSmoke Canada | The Ultimate ECig Electronic Cigarette HD Smoke()

  • http://www.facebook.com/mr.ray.anderson Ray Anderson

    The interesting idea would be apple to buy ARM who actually design their processors and drive a higher value of chip sales than Intel these days.

    • http://twitter.com/WalterMilliken Walter Milliken

      Too many anti-trust issues there, and not enough value. Apple already gets everything they need from ARM with the architecture license, which allows them to innovate on top of the standard ARM core designs. About the only additional value would be to cut off the competition from ARM, hence the anti-trust problem.

      • Walt French

        Yes, and buying ARM would merely create a vacuum for a replacement that all the other embedded devices would turn to—we talk about Apple’s reluctance to rely on a competitor, but surely it’d be far worse for a competitor to rely on Apple.

    • http://www.asymco.com Horace Dediu

      I don’t believe that ARM designs current processors for Apple. Apple’s A6 processors are designed in-house. See: http://en.wikipedia.org/wiki/Apple_A6

  • http://twitter.com/m4rkusr Markus Ramstedt

    Samsung today went on the record to say they’ll ship 510M handsets
    this year with 390M of them being smartphones
    (http://www.koreatimes.co.kr/www/news/tech/2012/12/133_127496.html).

    With the S III selling roughly at a run rate of 5M a month and the Note 2 at
    roughly 2M a month based on the numbers published by Samsung (these are
    likely the two key volume products with Exynos chips, Note II having
    Exynos engine also in the US) it’s actually pretty curious that
    something like 75% of Samsung smartphones actually ship with non-Samsung
    chipsets. This likely makes Samsung the biggest customer for companies
    like STE, Broadcomm and Marvell.

    If Samsung would move to in-house chipsets, many of the already struggling mobile chipset vendors would likely go belly up.

  • dmcdee

    Given the amount of analysis and research to which Apple is subject, including that by the redoubtable Horace, it is astonishing that there do not seem to be informed guesses as to what Apple is investing in. We are assured by Horace that the $10 billion or so is depreciable equipment – surely this would eventually be visible as orders on suppliers, physical deliveries and installations. Anyone any ideas?

  • Pingback: Apple Is Blasting A Multi-Billion Dollar Hole In Samsung’s Business (AAPL, SSNLF) | Digital Wealth()

  • Pingback: iNSIDER » Apple Is Blasting A Multi-Billion Dollar Hole In Samsung’s Business()

  • Pingback: Apple Is Blasting A Multi-Billion Dollar Hole In Samsung's Business -TekDefender()

  • dmcdee

    Further to my last as to the visibility of this scale of investment: Intel’s Fab 42 coming on stream this year will have cost a mere $5 billion all in, over two years, creating “thousands” of jobs. There must be a limit to what Apple can hide or camouflage.

  • http://twitter.com/matthewwanderer matthew

    “an estimated 226 million processors in 2012.”

    Apple is quickly closing in on demand requiring 1-million processors be manufactured per day. Does the capital expenditures guesstimated for fabs suggest that sort of capacity?

    That number has me questioning the current wisdom that Apple will cut ties with Samsung in the near future. Apple might be building fab(s) simply to ensure ample supply…which will have the secondary benefit of diluting their dependence on Samsung.

  • Pingback: Apple Is Blasting A Multibillion-Dollar Hole In Samsung's Business | Galaxy SIII()

  • Pingback: How will Samsung respond to the challenges from Apple and Google? | Decoding the new economy()