Watched the video twice and had some thoughts…

From friend-of-the-blog John Garner:

Good morning, Phil

Watched the video twice (not all of the 2nd run through) and have some thoughts on it.

This was by far the best overall presentation that I’ve seen from Apple at such events for several years.  While still not to Steve’s standard, it was markedly better than what I saw of the Developer Conference in the spring.  Much tighter and everyone seemed far better prepared.  Really enjoyed the introductory video – quite humorous.  Particularly noticeable improvement with Jeff Williams and Tim Cook.  And Phil Schiller wasn’t repeating himself for a change.

Was particularly impressed with the Apple Watch.  I’ve been kind of neutral on the whole device – saw some advantages and was impressed with the technology but there’s none on my wrist.  But I have both AFib and sleep apnea problems, and know a great deal about ECGs.  What Apple put into its new Watch is really quite amazing.  Throw in the new complications, and I will be buying one of these.

Nothing about the iPhones XS and XS Max particularly surprised me.  While they’re nice products, between the natural development of Apple’s technologies (and the leaks,) only the cameras and the speakers surprised or impressed me.  (See later comments on A12 chip.)  What’s happening with cameras and software blows me away with each generation.  The variable depth of field was amazing.  With the exception of telephoto lens, I can’t even imagine using an SLR or any of my other cameras now; I remember vividly what I used to lug around with me on field trips.

Apple also appears to have put huge efforts into its speakers this year both in the MacBook Pros and the iPhones.  It’s been well worth the effort and huge improvements.

The iPhone XR really surprised me, and I suspect will be a huge seller.  Apple made a truly smart marketing move here by differentiating largely on the screen and memory, and not on the available feature set.  While it’s not been said (and I reviewed the tech specs and the early technical reviews closely), I suspect that Apple/TSMC are binning the A12 chips and are using the slower ones in this phone.  It’s a smart move financially and from a product perspective.

There are several rather interesting implications from these announcements:

  1. TSMC’s chip yields on their new 7nm process node are much higher than anyone has expected at this time.  To see a 7nm A12 was no surprise to me as you know; to see this chip in all of the iPhone XSs and XRs definitely was.  The confidence that Apple is showing in TSMC’s capabilities with this level of commitment to the 7nm process node is quite amazing.  If there’s a risk point in Apple’s new iPhones, that’s exactly where it sits.  We’ll know soon from the lead times on iPhone deliveries.  Sales at very high volumes may be a challenge.  Is concern about chip supply why we didn’t see new iPads and MacBooks???
  2. The implications for Intel who can’t even get the chip yields from their 10nm process node to adequate levels two years after it was supposed to be fully operational are very, very dangerous.  As I said to you months ago, short Intel.  My opinion on that just got much, much stronger.
  3. The A12 chip design appears quite conservative.  I suspect that Apple did not want to push the design envelope hard with this chip, and covered its butt with a design that it could make on TSMC’s 10 nm node and get the same performance by increasing the clock speed.  The 15% performance increase is about exactly what you’d expect from shrinking an A11 chip design, and splitting the performance and power improvements that you’d get.  Apple’s next chip, likely the A12X will push performance on both the CPU and graphics sides, and that’s the one which I expect to power the next iPads and low-end MacBooks.
  4. It would be interesting to know which of ARM’s instruction sets Apple used in designing the A12, i.e., Cortex A76?  That’s geared to laptops and tablets.
  5. Don’t get too excited about the neural engine’s performance numbers.  They’ve achieved the huge performance gain by adding six more neural cores, and things are not exactly what they may appear with the performance numbers.  What’s the precision of the calculation???  However, Apple would not have used this many transistors if it weren’t getting now (or expected to in the future) something from the extra transistors in this design.
  6. There’s NO information on clock speed available on the A12 chips.  The performance numbers will become available over time as these iPhones are distributed, but the clock speed itself is really one of the key things that we need to know to assess the future paths of Apple’s chip designs.  I suspect that Apple’s continuing to use relatively slows speeds to boost battery life.


My take: I don’t know enough about chip yields to have a take.

To rewatch the video, click here.


  1. Gregg Thurman said:

    Thinking along the lines of the A-12 and how much more powerful it is over anything offered by the competition, I can see the iPhone XR being a deadly Android rival over the next two years. Yes, two years.

    That power drives some incredible technology on the iPhone, technology that is going to improve over time. It’s going to take at least two years for Qualcomm, Huawei, and Samsung to match Apple’s A-12 performance capabilities. Without A-12 like performance from Android product processors, Google is going to be hampered in improving Android. It’s a double-edged sword (hardware and software) undermining Android competitiveness.

    In that two year period the iPhone XR, powered by an A-12 processor, will be reduced (as is Apple’s practice) in price to $549. I see a rather large turnover of iPhone’s installed base over the course of the next three years, not to mention Android switchers, that will collectively drive significant unit sales growth.

    September 13, 2018
  2. Ken Cheng said:

    It will be interesting to see the chip speed of the A12 in the Xr and the speed in the Xs. If the writer is right, we’ll see some segmentation between slower A12s in the Xr and faster ones in the Xs. Makes sense to improve chip yields, since even the rejected ones for the Xs could be fast enough for the Xr. We’ll see when the benchmarks get done.

    September 13, 2018
  3. Fred Stein said:

    Hooray. Someone else sees the value of the A12.

    Apple quietly told the world, “We’re untouchable.”

    How did they (Apple and TSMC) get there – best chip in universe? “Incremental” that cursed incremental. Apple has incrementally improved their A series chips for years, each time pulling further ahead. Likewise, TSMC is the best foundry, making the 7nm chips for Qualcomm and Huawei. But Apple is at the head of TSMC’s queue. Samsung’s 7nm lags by about 6 months. Intel is way behind.

    More importantly, only Apple can optimize the full stack, as in all the specialized cores, the firmware, the low-level OS, and the higher layers like Core ML, ARKit, etc.

    The above helps understand the breadth of Apple’s iPhone options. While the Xs and Xs Max will own the top end, the 8 and the Xr model grab the next tier down. The 7 Plus with telephoto starting at $569, steals the show below $600. Apple’s exposure below $449 get filled by used and hand-me-down phones as 230M people upgrade over the next 12 months.

    Android will resemble hungry dogs fighting over scraps. iPhone installed base will continue to grow. It is not yet saturated.

    September 13, 2018
  4. Steven Noyes said:

    I think many people are selling the new Apple Watch short and underestimate the impact it will have this Christmas. I look at the Apple Watch, and like you, see it moving from a niche device to a near “must have” device for iOS owners. I see it driving near iPhone revenue within a few years.
    Okay, I will have to admit. I saw lots to like in the Apple Watch and the FDA approval was another shout out to competitors (like the 7nm A12) stating “we are years ahead of you”.

    September 13, 2018
  5. David Drinkwater said:

    “The A12 chip design appears quite conservative. I suspect that Apple did not want to push the design envelope hard with this chip, and covered its butt with a design that it could make on TSMC’s 10 nm node and get the same performance by increasing the clock speed.”

    This looks a little bit like the typical tick-tick chip cycle:

    “Tick”: present new architecture at current technology (manufacturing capability, typically expressed in critical dimensions: 1nm = 10 A =~ 3 silicon atoms in a row – although the don’t strictly form rows). With 7 nm being about 20 silicon atoms across, we are certainly challenging some physical limits.

    “Tock”: take today’s design and apply a new technology node to it. I don’t know what could possibly come next. 7 nm is already pretty phenomenal. That’s why we’re surprised to interpret that TSMC is having good yields.

    Anyway, with “tock”, you take how you understand circuit design and then manufacture it better.

    When the next “tick” comes around, you take what you’ve learned how to manufacture (the actual *making* of stuff and see what you can *design* better with those new capabilities.

    Then you “tock” again and learn to make your new designs better.

    Sometimes a new material can step in and change things in a big way, but I struggle to think where we are going to go next.

    Still, it seems like Apple has a very good partnership going with TSMC. That’s good for both of them.

    September 13, 2018
  6. David Emery said:

    The key problem for system speed once you’re near 7nm and run out of atoms 🙂 is not hardware, but software. It’s the ability to split up a problem across multiple cores. Those are areas of competence for both Apple and Google (but in somewhat different domains of application.)

    September 13, 2018

Leave a Reply