How TSMC Won Back Exclusivity With Apple for the A10 Chip in iPhone 7

Discussion in 'MacRumors.com News Discussion' started by MacRumors, Aug 10, 2016.

  1. JulianL macrumors 65816

    Joined:
    Feb 2, 2010
    Location:
    London, UK
    #101
    Potentially really exciting stuff but...

    Please, someone in Apple, this time stand up to Jony Ive (or whoever is really responsible for Apple's thinness obsession) and use any space savings inside the case to increase battery size rather than let the physical designers grab all the benefits from these sorts of technical innovations and use them to make the device even thinner. I think we must be at at least a 10 to one ratio now between people saying "I want longer battery life" vs people saying "I'd like the next iPhone to be even thinner".

    I'm still waiting for the iPhone when Apple decides to throw everything at battery life. Big efficiency improvements in components together with actually using extra space in the case gained from higher levels of integration to increase battery size really might at some point let Apple bump battery life 30-40% and maybe more between iPhone models without needing to roll out some new exotic battery chemistry (although if they could do that as well then even better).
     
  2. alo9000 macrumors newbie

    alo9000

    Joined:
    Aug 13, 2016
    #102
    That's because they poach TSMC's employees, some of which brought trade secrets and broke non-compete agreements.
     
  3. pkginstall Suspended

    Joined:
    Aug 5, 2016
    #103
    I would think he probably knows what he's doing, don't you?
    --- Post Merged, Aug 13, 2016 ---
    It's really not that difficult a concept to understand, how the chip is packaged and attached to the pcb. Google "BGA" - it's just a grid of solder balls which melt and fuse to matching pads on the logic board, not quantum physics.
     
  4. grkm3, Aug 13, 2016
    Last edited: Aug 13, 2016

    grkm3 macrumors 6502a

    Joined:
    Feb 12, 2013
    #104
    can you back that statement up? if you are talking about liang the guy that left tsmc in early 2009 and went to work for Samsung he was sued not Samsung

    they won in 2011 against the guy not Samsung and if you think his input from 2008 technology had anything to do with 14nm today you are delusional.tsmc was at 40nm when that guy was working there
     
  5. pkginstall Suspended

    Joined:
    Aug 5, 2016
    #105
    Maybe you're not aware of how design processes work at Apple? The vision of the IDg is realised in many, many models and prototype builds, one of which is finally chosen, and then the engineers are told to make everything fit inside that design, and if you have a competent engineering team, they WILL manage it.

    In the old days, Apple was engineering-driven, meaning that the "design" was a bunch of electronics nerds building circuits and then saying to IDg "okay, put a skin on this" (wrap it up and prettify it) - back to front for a design-driven company, so now design is how Ive and IDg envision it, instead of the engineers dictating the size and dimensions of the form.

    It's completely right that the visionary gets what he envisions, and doesn't get constrained by anything (except physics, of course - no escaping that!)
     
  6. smalm macrumors newbie

    Joined:
    Aug 26, 2014
    Location:
    Civitas Tautensium, Agri Decumates
    #106
    He was not just some guy, he was R&D senior director.
    <<starting with 45nm to 28nm, the difference between Samsung's and TSMC's technologies narrowed, the report found. "The 16nm and 14nm FinFET products that both companies will mass produce this year were even more alike," the report indicated. "It could be hard to tell (if the product) came from Samsung or TSMC if only structural analysis is used.">>

    Pure coincidence, I'm sure :cool:
     
  7. grkm3 macrumors 6502a

    Joined:
    Feb 12, 2013
    #107
    Samsung was never sued.the guy left in 2009 and had nothing to do with Samsung.its real easy to make a click bait headline for page hits.if Samsung used any of there patented technology tsmc would if brought them to court.
     
  8. mi7chy macrumors 603

    mi7chy

    Joined:
    Oct 24, 2014
    #108
    Apple went with TSMC because it's cheap and not because it's good.
     
  9. astroboy888 macrumors member

    Joined:
    Aug 9, 2013
    #109
    Samsung's semiconductor manufacturing technology is funded by their own semiconductor business in memory and LCD panel, which is very lucrative for them. Samsung spends about $12B USD a year on manufacturing equipment. (TSMC spends about the same per year) But if you separate out Samsung's contract manufacturing business only. Samsung makes about ~$3Billion revenue per year far less than what TSMC makes ($26Billion USD per year). But of course Samsung makes up the loss by using profit from other business units.
    --- Post Merged, Sep 29, 2016 ---
    TSMC did sue the guy who was convicted of corporate espionage. This guy was the director who was leading TSMC's 16nm development for many years. Samsung arranged for him to by pass his non compete agreement by giving him a teaching job at a Samsung sponsored university to current Samsung semiconductor engineer employees for two years. Allegedly transferred TSMC's sensitive technology and technology in this manner. The fact is there is absolutely no way you can bring a 14nm technology online within 3 years without outside help. TSMC had been working on it FinFet for 15 years and 16nm for 5-7 years previously with this guy at the helm. Intel worked on theirs for 10 years. So it was pretty obvious to industry insiders what had taken place.

    TSMC chose not to sue Samsung. Instead TSMC simply reminds potential customer the threat and the risk of using Samsung's 14nm process. If TSMC does decide to pursue a legal route, it might result in their products being banned from major world markets. This risk alone was effective enough to persuade many chip companies to continue to contract services with TSMC.
    --- Post Merged, Sep 29, 2016 ---
    Apple went with TSMC because the yield was higher and the device resulted in lower power. This fact was confirmed through the Apple A9 "Chipgate" incident. Apple ended up paying a higher price for TSMC than they did Samsung. Samsung lowered the price when its 14nm process did not meet minimal yield requirement that Apple asked for.

    Apple went back to TSMC and ask for the lower price. TSMC didn't budge. Basically saying to Apple, "you get what you paid for".

    For the record Apple pays TSMC $3.6 Billion per year or about 7% of TSMC total business. This is more than entire Samsung contract manufacturing revenue on a per year basis.
     
  10. tooltalk, Sep 30, 2016
    Last edited: Sep 30, 2016

    tooltalk macrumors 6502

    Joined:
    Jan 15, 2015
    Location:
    NY, NY
    #110
    That's an interesting insight. So what is Samsung LSI's capacity and what is their utilization rate after Apple's move to TSMC? I'm guessing that, based on your comment about Samsung's loss and having to make up the loss from other business units; it's not very high, their new customer Qualcomm not withstanding. But on the other hand, I've always thought that Samsung's customer/technology sharing agreement with Glofo, Common Alliance partner, meant that they didn't have enough capacity to meet customer's demand. Samsung must be in a dire financial condition absence Apple's contract. /s

    Further, most of Samsung's $12B Capex is going to Samsung Semi in general, not to the LSI division in particular, though that capex figure you cited seems to be a bit off. Or are you saying that Samsung LSI alone spends $12B capex yearly, compared to TSMC's similar amount, but the LSI only generates $3B, compared to TSMC's $26B? That seems like irrelevant, incorrect, or confusing at best (perhaps deliberately?)

    [​IMG]


    Also with respect to Samsung's 14nm, how much of it do you think is directly from TSMC vs IBM from Common Alliance? Can a single person, though I suspect there is plenty of poaching in the industry, out of hundreds, if not thousands working on various components, make that much difference? My ex-roommate (PhD in EE/Physics) worked at IBM (first in Vermont, then later Upstate) and now at Glofo, so I know that IBM trained many dozens of Samsung engineers from South Korea to get them up to speed at their Upstate campus many years ago (as part of Common Alliance). But he was never privy to the management decision or sales contract details -- and sounded quite oblivious when asked about these.

    Just out of curiosity, do you have actually have contacts at TSMC, IBM/Glofo, Samsung or Apple to substantiate anything you've said (and I don't mean your online friends at SemiWiki) or are you one of those arm chair generals who scan headlines all day long?
     
  11. tooltalk macrumors 6502

    Joined:
    Jan 15, 2015
    Location:
    NY, NY
    #111
    I believe poaching is quite common in that small industry, as it is quite common in that software industry I'm in. So what lead you to believe that one person out of hundreds and thousands of engineers and researchers who spent several years and billions of dollars developing the processing had all to do with Samsung's 14nm ?
     
  12. astroboy888, Oct 5, 2016
    Last edited: Oct 5, 2016

    astroboy888 macrumors member

    Joined:
    Aug 9, 2013
    #113
    CAPEX is an catch all index combines all semiconductor sales and manufacturing service. It is useful as an financial indicator for the overall semiconductor industry, but doesn't tell the story of individual segment within the industry itself. Selling semiconductor and provided fab services to 3rd party are separate businesses segments. TSMC is a "pure play contract manufacturing service". This means TSMC does not sell chips, but only provide addon service to 3rd party chip designers such as Apple and Nvidia (and Samsung and Intel). On the other hand, Samsung and Intel has separate business units that sells chips AND provide manufacturing services to 3rd party chip designs. If we look at the latter, contract manufacturing services alone, TSMC is by far the largest.

    Regarding your question. If you take the history of Samsung production on A7 CPU (28nm) as an exmaple, Apple's orders occupied 50% of Samsung's total capacity. So I would venture to guess that it would be about the same had Apple gone 100% into Samsung 14nm. In comparison, ALL Apple Silicon needs occupied around maybe up to 4-5% of TSMC's total yearly output. (But accounts for 7-8% revenue) Apple's business with TSMC covers a lot more than the A-series CPU, peripheral chip sets and sensors were purchased from TSMC or TSMC joint ventures. Similarly Apple's depends on Samsung for RAMS, SSDs, and LCD panels. But certainly A-series CPU manufacturing is the crown jewel and bragging right for bleeding edge technology from a marketing perspective.

    The Common Platform was created for the sole purpose of allying against the domination of TSMC about 18 years ago. IBM had always been pushing planar FD-SOI (basis for Samsung 28nm and 20nm node). It is a direct competitor of FinFet technology, so FinFet was not an area of IBM's expertise. FinFet was invented by UC Berkeley Professor Chenming Hu, who also served as TSMC CTO from 2001-2003. Both TSMC and Intel were firmly on FinFet's camp early on and started working on it from late 1990s. Flash forward, Samsung, seeing problems with FD-SOI scaling below 20nm, "developed" its own 14nm FinFet; therefore Common Platform members such Global Foundary and Taiwan's UMC had to license it from Samsung. Incidentally Global Foundary is now pushing its own 12nm planar FD-SOI as a solution to compete against TSMC 16nm technology node, Samsung (and its own) 14nm FinFet. It is unknown if Samsung will follow suit. (Probably not).

    Non-compete agreement is common in any technology segment not just software. Liang Mong-Song, was not just an employee, he was a senior director of R&D and the technical lead for TSMC 16nm technology development. He was in possession of all TSMC 16nm trade secrets when he was targeted and poached from TSMC. He had signed a non-compete agreement not to work in the Semiconductor industry for an unspecified number of years To get around the agreement, Samsung arranged for him a teaching job in South Korea at an Samsung sponsored university teaching Samsung employees. The entire electrical engineering department was practically on Samsung's payroll. According to the court papers, Samsung benefited in its 28nm node and possibly speed up development of Samsung's 14nm technology. It should be noted that Mr. Liang was only convicted of passing 28nm trade secrets through his teaching position. However, industry insiders pointed out that Samsung brought its 14nm FinFet technology online within 5 years having no experience previously, whereas both Intel and TSMC worked on FinFet for more 15 years starting late 1990s. It is extremely difficult to accomplish such a task without "outside" help. Mr. Liang was not the only one who was poached. Many Intel, Global Foundry, and TSMC employees were poached by Samsung in its efforts to catch up with the likes of TSMC and Intel.

    Do you know who Wernher von Braun and Qian Xuesen was? von Braun was the former Nazi German rocket scientist who built NASA's space program and designer of Saturn Rockets. He was chiefly responsible for giving USA jet propulsion technology used in rockets and military fighters. Qian Xuesen, former Cal Tech professor and US rocket scientist, first director of the JPL, was turned by Russia and China, traded by the US Army for 500 POW during the Korean War. Qian gave China ballistic missile technology, the space program, and the atomic weapon. Yes, one person can and does make that much difference.
     

Share This Page