Apple iPhone supplier TSMC on track to ship 10nm chips in this quarter, 'expand rapidly'

Posted:
in iPhone
Apple's main mobile processor manufacturer, TSMC, is reportedly set to begin commercial shipments of 10-nanometer chips in the next month -- likely supporting claims that the "A11" processors in upcoming iPhones will use the technology.




Shipments of 10-nanometer chips should grow rapidly in the second half of the year, co-CEO Mark Liu confirmed at a company forum according to DigiTimes. Apple is typically expected to ship three new iPhones in the fall, including an "iPhone 8" as well as an "iPhone 7s" and "7s Plus."

Previous rumors suggested that TSMC had won orders to make a 10-nanometer "A11," and indeed other manufacturers such as Samsung are due to use the technology this year.

The processor in the iPhone 7 and 7 Plus, the A10 Fusion, is built using TSMC's 16-nanometer FinFET process. Shrinking die size can be money-saving for manufacturers, and beneficial to end users as well, since it often results in more efficient chips with less power consumption than would've otherwise been possible.

TSMC is meanwhile said to be planning "risk production" of 7-nanometer chips in the March quarter, switching into mass production in 2018 -- possibly in time for that year's iPhones. 5-nanometer risk production is already slated for the first half of 2019.

While the "iPhone 7s" and "7s Plus" will likely be evolutionary upgrades, the "iPhone 8" is expected to feature a 5.8-inch, edge-to-edge OLED display, of which 0.7 inches will be dedicated to virtual buttons, replacing the physical home button Apple has had since 2007. The phone may also have 3D facial recognition and/or iris scanning.
«1

Comments

  • Reply 1 of 23
    wood1208wood1208 Posts: 2,905member
    Nothing new here. 7nm production would be interesting.
  • Reply 2 of 23
    Wow, they are really moving fast. Desktop CPU have been shrink from 45nm (2008) to 32nm, to 28nm, to 22nm, to 14nm for a long time. It was just 2011/2012 that Apple uses the 45nm technology for the A5 chip. Now we are talking about 10nm this year and 7nm in 2018? I know this was the roadmap, but still, wow!
    patchythepirateStrangeDaysSoliwatto_cobra
  • Reply 3 of 23
    FatmanFatman Posts: 513member
    Wow - this is big - and sooner than expected. Intel has been struggling to achieve this milestone (in large quantities). Customers will benefit from longer battery life and faster procs! Hats off to the designers and engineers that make it possible.
    watto_cobra
  • Reply 4 of 23
    It's interesting how processor manufacturers used went from focusing on the development of faster and faster processors, to now focusing on energy efficiency. Once they hit the 5nm mark, I can't imagine there will further improvements without some significant shift in overall processor design.
    watto_cobra
  • Reply 5 of 23
    Help me understand this.  An average strand of DNA is approx 7nm across.  How can chips be built in this fashion?  Also doesn't quantum tunneling come in to play here?
    sfjohntallest skil
  • Reply 6 of 23
    flaneurflaneur Posts: 4,526member
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    edited February 2017
  • Reply 7 of 23
    wizard69wizard69 Posts: 13,377member
    wood1208 said:
    Nothing new here. 7nm production would be interesting.
    Nothing new but the potential for iPads and iPhones is very exciting.    Predicting what Apple will ship is impossible but with 10 nm we will have the possibility of quad cores of very high performance for mobile devices.   And of course expanded GPU performance.   Basically Apple could ship laptop class performance in a cell phone.   Not just low end laptop performance either, they may be able to match a 13" MBP on some workloads.
    sfjohnwatto_cobra
  • Reply 8 of 23
    wizard69wizard69 Posts: 13,377member
    It's interesting how processor manufacturers used went from focusing on the development of faster and faster processors, to now focusing on energy efficiency. Once they hit the 5nm mark, I can't imagine there will further improvements without some significant shift in overall processor design.
    Energy efficiency was actually demanded by Apple.   They pushed Intel hard here but Intel has stumbled bad, thus we are starting to see other manufactures eclipse Intel with respect to power usage and feature size.    

    The shift will likely come with above to Graphene based processors.   At some point feature size shrinks will end and we will needs something to replace silicon.   When this happens we will see a rush back to performance and clock rate increases as that will be the only way forward to improve conventional processors.   
    tmay
  • Reply 9 of 23
    melgrossmelgross Posts: 33,510member
    jcdinkins said:
    Help me understand this.  An average strand of DNA is approx 7nm across.  How can chips be built in this fashion?  Also doesn't quantum tunneling come in to play here?
    Yes to both. But it needs reminding that these days, more so than in the past, these process jumps aren't altogether what they seem. It's estimated that both Samsung's and TSMC 10nm is about equal to Intel's 14nm. Not all features on a chip are of the process size, many, if not most, are much larger.

    quantum effects have been noticed since 120nm. At 90nm they became much larger than anticipated, and so ended the march to higher speeds. Intel, IBM and AMD were predicting that CPUs would reach 15, and even 20 GHz before now. As we know, that never happened. Instead, they went to multiple cores.

    quantum effects will take over at about 5nm. Some companies are saying 5nm will be coming in a few years, but a number of chip,experts are doubting it. At that width, a line is only about 12 atoms wide.
    Solisfjohnwatto_cobrafrantisekanalogjack
  • Reply 10 of 23
    melgrossmelgross Posts: 33,510member

    flaneur said:
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    Nah. 5nm seems to be something that might be possible, but might not. I read an article that said that 3nm was possible, but I highly doubt it. It has less to do,with the masking technology than to physics saying that many electrons will tunnel out of the feature as fast as it's put in. Don't worry about what Digitimes has to say about this.
    SpamSandwichwatto_cobra
  • Reply 11 of 23
    What part of a chip does the 7nm refer to?
    SpamSandwich
  • Reply 12 of 23
    ciacia Posts: 247member
    melgross said:

    flaneur said:
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    Nah. 5nm seems to be something that might be possible, but might not. I read an article that said that 3nm was possible, but I highly doubt it. It has less to do,with the masking technology than to physics saying that many electrons will tunnel out of the feature as fast as it's put in. Don't worry about what Digitimes has to say about this.
    The real question is that we clearly are approaching a wall. Where do we go from here? In 15-20 years what will be the solution if we can no longer shrink? More cores? Quantum computing?
  • Reply 13 of 23
    When Intel gives a scale, say 14nm for their chips you know what you are getting. TSMC & other chip manufacturers use their numbers for marketing purposes only, and it doesn't really tell you anything about the actual size of their chip when comparing them to Intel. It's great that Apple will be getting chips that are smaller. TSMC 10nm is very likely somewhere near an Intel 14nm in size. When and if Intel is able to produce an actual 7nm chip, that's going to be it for size reductions. Below that quantum tunneling will force chip makers to look for other ways to Increase their speed. I imagine we'll start to see the number of cores rise.
  • Reply 14 of 23
    What part of a chip does the 7nm refer to?
    Nothing. It is mostly marketing now that comes from a standard terminology from International Technology Roadmap for Semiconductors (ITRS). For example, according to Wikipedia, the 14 nm technology fabrication node has a variety of feature sizes (these are Intel's numbers): Transistor Fin Pitch 42 nm, Transistor Fin Width 8 nm, Transistor Fin Height 42 nm, Transistor Gate Pitch 70 nm and Interconnect Pitch 52 nm.

    If you want more info, this is a good site https://en.wikichip.org/wiki/14_nm_lithography_process.


    Edited to add wikichip.org site.
    edited February 2017 ireland
  • Reply 15 of 23
    cia said:
    melgross said:

    flaneur said:
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    Nah. 5nm seems to be something that might be possible, but might not. I read an article that said that 3nm was possible, but I highly doubt it. It has less to do,with the masking technology than to physics saying that many electrons will tunnel out of the feature as fast as it's put in. Don't worry about what Digitimes has to say about this.
    The real question is that we clearly are approaching a wall. Where do we go from here? In 15-20 years what will be the solution if we can no longer shrink? More cores? Quantum computing?
    Quantum computing might happen but it isn't going to replace our current general purpose computing model. There are other materials that show promise but they are probably 15-20 years out maybe even more. It is likely that Moore's law will stall at least for a while very soon.
  • Reply 16 of 23
    flaneurflaneur Posts: 4,526member
    melgross said:

    flaneur said:
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    Nah. 5nm seems to be something that might be possible, but might not. I read an article that said that 3nm was possible, but I highly doubt it. It has less to do,with the masking technology than to physics saying that many electrons will tunnel out of the feature as fast as it's put in. Don't worry about what Digitimes has to say about this.
    1. What, me worry? 2. DigiTimes has nothing to do with the matter under discussion; they were merely reporting (form Taiwan, a bit closer to sources than New York) a reasonable possible factoid about TSMC's technology roadmap. 3. The knee-jerk dismissal around here of anything DIgiTimes carries is tiresome, and is itself a cliché. 

    I'd be more interested in informed commentary about IC production technology that's designed to find the actual quantum tunneling limits empirically, which is what TSMC may be doing, in defiance of what theory predicts. Or not.
    edited February 2017
  • Reply 17 of 23
    irelandireland Posts: 17,798member
    So what processor does 7s get?
  • Reply 18 of 23
    irelandireland Posts: 17,798member
    jdb8167 said:
    cia said:
    melgross said:

    flaneur said:
    I wonder if there's anything beyond 5nm too. This DigiTimes story says 5nm requires a shift to Extreme UltraViolet (EUV) for the lithography.

    http://www.digitimes.com/news/a20170223PD212.html

    Might they discover a new round of shrinking possible at those wavelengths?
    Nah. 5nm seems to be something that might be possible, but might not. I read an article that said that 3nm was possible, but I highly doubt it. It has less to do,with the masking technology than to physics saying that many electrons will tunnel out of the feature as fast as it's put in. Don't worry about what Digitimes has to say about this.
    The real question is that we clearly are approaching a wall. Where do we go from here? In 15-20 years what will be the solution if we can no longer shrink? More cores? Quantum computing?
    Quantum computing might happen but it isn't going to replace our current general purpose computing model. There are other materials that show promise but they are probably 15-20 years out maybe even more. It is likely that Moore's law will stall at least for a while very soon.
    If this law is stalling then is not a law, just a statement that was true for a short while. Humans, always trying to build a box to put themselves in.
    edited February 2017 calianome
  • Reply 19 of 23
    TSMC at 10 nm is more or less equivalent to Intel at 14 nm. TSMC wanted to skip over 10 nm, but I suspect that Apple may have pushed them. TSMC does plan on moving to 7 nm in fairly quick fashion where they would take the lead from Intel if Intel continues to run into difficulty at 10 nm. Intel is trusting their Israeli plant to deliver but we will have to see whether they are able to before TSMC gets down to 7 nm. TSMC has grand plans at 7 nm also. 

    TSMC has plans to shrink their nodes further, but a healthy dose of skepticism comes into play. 

    One thing is for certain, Intel's process lead has essentially vanished. And they are now being pushed with respect to CPU design by Apple's A series SOCs, never mind the upcoming AMD Ryzen. 

    The most interesting aspect to me is whether the next generation iPad (pro) will be released in March and whether the A10X chip is built on 10 nm. 
    ksec
  • Reply 20 of 23
    foggyhillfoggyhill Posts: 4,767member
    melgross said:
    jcdinkins said:
    Help me understand this.  An average strand of DNA is approx 7nm across.  How can chips be built in this fashion?  Also doesn't quantum tunneling come in to play here?
    Yes to both. But it needs reminding that these days, more so than in the past, these process jumps aren't altogether what they seem. It's estimated that both Samsung's and TSMC 10nm is about equal to Intel's 14nm. Not all features on a chip are of the process size, many, if not most, are much larger.

    quantum effects have been noticed since 120nm. At 90nm they became much larger than anticipated, and so ended the march to higher speeds. Intel, IBM and AMD were predicting that CPUs would reach 15, and even 20 GHz before now. As we know, that never happened. Instead, they went to multiple cores.

    quantum effects will take over at about 5nm. Some companies are saying 5nm will be coming in a few years, but a number of chip,experts are doubting it. At that width, a line is only about 12 atoms wide.
    Sometimes, the actual limits are economics, not even technical.
    They could possibly make it at 5nm, with all sort of caveats about the actual reality of this (wrapping them to reduce leakage at junction, stacking, multipass UV, whatever), but if it costs 10 times the cost to develop it with really shit yields, is it actual worth while to do so.

     At that point, previously non economically viable techs will in fact emerge as progress stalls with current tech.
    They'll start expensive and niche and eventually spread to more mainstream use.
    ksec
Sign In or Register to comment.