Comments Locked

24 Comments

Back to Article

  • Hyperlite - Friday, September 20, 2013 - link

    I know it's pretty standard practice in the industry, but i still find it infinitely ironic.
  • ArthurG - Friday, September 20, 2013 - link

    Apple has no factory, what you think they can do ? They rely on others for all their hardware so they just try to find the best deal in everything...
  • Kevin G - Friday, September 20, 2013 - link

    The Intel foundry deal rumor has cropped up again?

    I get it why the manufacturing side of Intel would want this but the x86 of the company has to be irate about the idea of supporting a direct competitor. Remember, Intel is keen to push x86 into smart phone via Atom and manufacturing chips for a smart phone competitor doesn't bode well for Intel's image. Not to say that Intel couldn't be successful in both selling x86 silicon in mobile devices while simultaneously fabing chips for Apple's iPhone, it doesn't make for good marketing when you're trying to really break into a market.
  • Bob Todd - Friday, September 20, 2013 - link

    True, but fabs are expensive assets. I remember reading something earlier this year speculating that their fabs were running around 60% utilization. Their process advantage is still large, and I imagine Apple could do some pretty impressive stuff at 14nm/7nm. There's obviously a ton of variables to think through like a possibly uncomfortably small margin for Intel. However I'm sure the chance to have a crack at those nodes well before their competitors is at least intriguing to Apple and others.
  • ltcommanderdata - Friday, September 20, 2013 - link

    If Intel has unlikely to get an x86 chip onto an iOS device in the near or even medium term, which seems pretty certain now that Apple is investing for the future in 64-bit ARM, and iOS devices still represent significant volumes even if it is behind Android, then Apple really isn't a direct competitor to Intel chips. It's not like Apple not using Intel chips in their iOS devices is going to drive a lot of customers away from iOS. iOS devices will continue to sell and sell well regardless of whether Intel x86 chips are in there or not. Fabricating Apple's chips would be an additive business for Intel rather than leaving a segment of the market untouched.
  • Kevin G - Saturday, September 21, 2013 - link

    There is of course the threat that Apple could migrate the Mac line to ARM in the future. Apple did state that the A7 was 'desktop' class which I see as indicative of where the Mac platform is going to go in the long term.
  • MartinT - Friday, September 20, 2013 - link

    Well, there is the old adage about Intel being primarily a process development and production company that just happens to also keep a design house around, mostly to generate the revenue that fuels development of the next node.

    Manufacturing is certainly the one part of the chain Intel is most dominant at, and has been for many a year.

    Once they ran out of in-house transistors to print on their shiny wafers, you saw them restarting their foundry business, and unless x86 suddenly replaces ARM in the mobile space, you'll only see this continue. Those FABs have to be fed.
  • errorr - Friday, September 20, 2013 - link

    Why couldn't intel offer up an n-1 fab strategy. Their process advantage is heads and shoulders beyond even TSMC let alone everyone else. We see TSMC start bulk 20nm in Feb. Assuming low yields early on and intel having 14nm up an running aroind the same time why couldn't we see intel offer to do volume 22nm finfet for the right customer?

    Most of the so called excess capacity that intel has is figuring out how long they can continue to monetize the older nodes.
  • beginner99 - Sunday, September 22, 2013 - link

    Yeah i wonder about this too. n-1 would make so much sense because almost everything is now integrated and hence chipsets are also on the same node. So factories get unused much faster. One can dream about NV/AMD GPUs made by intel 22/14 nm. :D
  • michael2k - Sunday, September 22, 2013 - link

    You are confused. They will be ramping up their high power 14nm process in Feb, their low powered process won't be up for another 12 months at least. If they were to go N-1 then they would be offering 22nm this time next year, at the same time TSMC will be running their 20nm process.
  • ErikZachrisson - Friday, September 27, 2013 - link

    TSMC has the advantage of being used to having multiple customers, thus they have very clear design rules. Intel on other hand has not such a workflow, having almost only internal customers.

    This makes porting designs to intel fabs hard and costly.
  • YawningAngel - Saturday, September 21, 2013 - link

    Actually, I can see fabbing for Apple potentially working to sell Atom chips to Android manufacturers. If you have to complete with the iPhone for power and SoC integration, you need a chip being fabbed on the same process. If Atom is the only such chip, it gives Intel more chance of grabbing market share from Qualcomm.
  • Kevin G - Saturday, September 21, 2013 - link

    The would require Atom to have a distinct performance as a well as a performance-per-watt advantage over what Apple can design. At this moment, Apple's A7 is very, very impressive in these areas. What Silvermont SoC's offer isn't enough to make the leap today. There is also the factor in that handset/tablet designers will be at the mercy of what Intel decides to put into Atom where as the ARM market is much of diversified in what SoC's are offered.

    I do see other SoC designers wanting to utilize Intel's fabs for their own SoC parts. Intel's Quark strategy appears to be playing into those desires if they want an x86 CPU put into an SoC.

    Between AMD's willingness to design SoC's for companies with outside IP and Intel's Quark strategy, it appears that handset/tablet designers have modern mobile and legacy desktop convergence on the horizon in some capacity. The only way to put legacy desktop support into your pocket is through x86 and Windows.
  • SodaAnt - Monday, September 23, 2013 - link

    Intel does have another advantage though, and that's process maturity. They can manufacture their own chips exclusively for a year or a year and a half at the smallest process, then let apple make theirs after that, and they'll still probably both ahead of the curve both in die yields and still be a bit ahead of TSMC.
  • Laststop311 - Saturday, September 21, 2013 - link

    Yea, I don't see how intel producing apples chips would negatively impact them in any way. Apple does all the work designing them. All intel has to do is manufacture the design apple gives them. It would be a great way for intel to utilize the unused capacity. Small margin of profit is better than under utilized capacity right?
  • Laststop311 - Saturday, September 21, 2013 - link

    it's actually a complete win for everybody. Apple gets to be happy and cut Samsung out of the picture while at the same time getting intels fully mature award winning 22nm finfet process for low power designs. Intel will have tons of 22nm capacity becoming available shortly when they move the bulk of their chips to 14nm. They are gonna have to do something with all that 22nm capacity. Even though they want x86 in all phones I just don't see how making apples Arm ISA chips would change anything as far as that is concerned. And even a year from now intels low power version of 22nm will still be cutting edge compared to all other ARM manufacturers.
  • mytakeismine - Saturday, September 21, 2013 - link

    Intel gets paid twice for their chips, once for the design and once for the manufacturing.

    When producing for Apple it would only get paid once and that chip is in direct competition. Now Intel would love to have the same deal with the Mac Air where they pay premium for an Intel chip and then get the chip a few months early
  • ananduser - Saturday, September 21, 2013 - link

    So...designed by ARM, modified/customized/enhanced by Apple, manufactured by Samsung. I think it's safe to say that the title of the 1st 64bit ARMv8 SoC goes to all of the above.
  • yhselp - Saturday, September 21, 2013 - link

    I feel this is more of a first step of Intel getting Apple on-board x86 for iOS. ARM is nowhere near (and might never be) the notebook/desktop segment and Apple won't stop producing such devices for years to come (if ever); they are however bringing deeper integration across all product lines and having the same architecture would help a lot. Given that Intel seems to have the superior architecture and certainly has the most advanced manufacturing process make them quite appealing.

    So unless something along those lines changes it appears that x86 and iOS are destined to meet, sooner or later. The fact that Apple are investing in ARM development isn't surprising, but doesn't necessarily mean that it's impossible for them to drop this altogether - they certainly have the vast resources to make such sudden shifts. Remember, Apple is a hardware/software company forced into microprocessor design because the situation necessitates it, while Intel is a microprocessor company proper; what this means is that Intel has the biggest interest to get into a market they missed and Apple wouldn't mind making more money by not developing their own architectures. Of course, you never know.

    P.S. What you guys say about Intel feeding then-old 22nm fabs with Apple SoCs make a lot of sense, but you might have missed the last sentence from this 'story' -- "Not to mention that the deal that was being worked on wasn't for 22nm."
  • Bob Todd - Saturday, September 21, 2013 - link

    "Apple is a hardware/software company forced into microprocessor design because the situation necessitates it"

    I don't think that's really accurate. Apple could have continued to use SOCs from other vendors like Qualcom/Nvidia/etc. The situation doesn't necessitate them bringing CPU design in-house. However, if you look at the revenue breakdown for their product mix, it's extremely heavy on the mobile side. I see their acquisitions and work thus far with A6/A7 as a way to have more control over their own destiny.

    And while I wouldn't really be shocked if Apple went x86 for iOS in another generation or two (when Intel is at 14nm/7nm), it also wouldn't really surprise me if they went the opposite direction. In several years we could see an Apple A11 powering Macbook Airs. It would be a huge undertaking for all 3rd party software, but it's not like Apple hasn't switched architectures before when it was to their benefit. If they could cut Intel CPUs out of their BOM and bank all of the increased margin between a $30 ARM part and a multi-hundred dollar Intel one they'd laugh all the way to the bank. That's easier said then done obviously since it's not just the CPU cores to worry about for their non-touch lineup. There's PCIe/Thunderbolt and a whole ton of crap in the chipset they'd need alternatives for as well.
  • vision33r - Saturday, September 21, 2013 - link

    Intel has zero interest in apple chip business unless it includes x86 adoption. They are not bleeding red and their business models depends on their x86 designs. Arm is a competitor there's no way intel would fab competing arm chips.
  • Qwertilot - Sunday, September 22, 2013 - link

    Apple really isn't a competitor though - very minimal chances of them using anything else for iOS and even harder imagining them licensing them outside - so they might. Now Qualcom....

    If Apple did get forced into making chips then it was by the enormous revenues of iOS. Just so big and profitable that there's enough scale to get a very good monetary pay back on doing it. The safety/independence is probably even more important mind. Same reasons that Samsung are likely to keep making chips I guess.

    I dunno if they'll ever switch the macs over or if they'll just enjoy menacing Intel with the idea :) Maybe if/when processors become outright commodities.
  • ancientarcher - Wednesday, September 25, 2013 - link

    If I could have a penny for every time someone said that Apple won't move its notebook line to its own ARM chips!! Have you seen the CPU/GPU benchmarks for the A7 before saying that ARM chips are nowhere near x86 designs??

    The A7 chip (on a 1.5W TDP) is 40% of the CPU performance of the ivy-bridge core i7 chips (at 35-45 W TDP) with graphics performance is nearly 50% of the Intel chips (please see Geekbench 3.0 results that Anand posted). And this, in a 28nm planar process node versus Intel chips at 22nm trigate. Now, what happens when the ARM ecosystem foundries move to 20nm next year and 14/16nm trigate in 2015? Don't you think that if Apple chip designers can bring out a 64bit low power ARM chip, they will also be able to do a higher energy (6-8W TDP) more powerful chip for its notebooks?? Why should apple want to pay $350-$450 for an Intel chip versus $20-30 for chips of its own design, particularly when its own design chips will be better across all dimensions - power consumption, CPU and GPU power???

    About legacy, Apple has done it before (from power to x86) and they will do it again (from x86 to ARM).

    Intel apologists need to think where things will be in a year's time
  • bobbozzo - Sunday, September 22, 2013 - link

    Hi, please expand or link acronyms; e.g HK+MG = "28nm Low-Power High-K Metal Gate Logic".

    thanks

Log in

Don't have an account? Sign up now