By means of introduction to my coverage of Apple’s 2024-so-far notable news, I’d like to share the amusing title, along with spot-on excerpts from the body text, from a prescient piece I saw on Macworld yesterday. The title? “Get ready for another Apple meeting that could have been an email”. Now the excerpts:
Apple started running virtual press events during the pandemic when in-person gatherings made little sense and at various times were frowned upon or literally illegal. But Apple has largely stuck with that format even as health concerns lessened and its own employees were herded back into the office.
Why is that? Because virtual events have advantages far beyond the containment of disease. Aside from avoiding the logistical headaches of getting a thousand bad-tempered journalists from around the world to the same place at the same time, a pre-recorded video presentation is much easier to run smoothly than a live performance…
Nobody cringes harder than me when live performers get things wrong, and I absolutely get the attraction of virtual keynotes for Apple. But it does raise some awkward existential questions about why we need to bother with the elaborate charade that is a keynote presentation. What, after all, is the point of a keynote? If it’s just to get information about new products, that can be done far more efficiently via a press release that you can read at your own speed; just the facts, no sitting through skits and corporate self-congratulation.
Is it to be marketed by the best hypemen in the business? If that’s really something you want, you might as well get it from an ad: virtual keynotes give none of that dubious excitement and tribalistic sense of inclusivity you get with a live performance. And we’ve even lost the stress-test element of seeing an executive operating the product under extreme pressure. What we’re left with is a strange hybrid: a long press release read out by a series of charisma-free executives, interspersed with advertisements.
I said something similar in my coverage of Apple’s June 2023 Worldwide Developer Conference (WWDC):
This year’s event introductory (and product introduction) presentation series was lengthy, with a runtime of more than two hours, and was also entirely pre-recorded. This has been Apple’s approach in recent years, beginning roughly coincident with the COVID lockdown and consequent transition to a virtual event (beginning in 2020; 2019 was still in-person)…even though both last- and this-years’ events returned to in-person from a keynote video viewing standpoint.
On the one hand, I get it; as someone who (among other things) delivers events as part of his “day job”, the appeal of a tightly-scripted, glitch-free set of presentations and demonstrations can’t be understated. But live events also have notable appeal: no matter how much they’re practiced beforehand, there’s still the potential for a glitch, and therefore when everything still runs smoothly, what’s revealed and detailed is (IMHO) all the more impactful as a result.
What we’ve ended up with so far this year is a mix of press release-only and virtual-event announcements, in part (I suspect, as does Macworld) because of “building block” mass-production availability delays for the products in today’s (as I write these words on Tuesday, May 7) news.
But I’m getting ahead of myself.
The Vision Pro
Let’s rewind to early January, when Apple confirmed that its first-generation Vision Pro headset (which I’d documented in detail within last June’s WWDC coverage) would open for pre-orders on January 19, with in-store availability starting February 2.
Granted, the product’s technology underpinnings remain amazing 11 months post-initial unveil.
But I’m still not sold on the mainstream (translation: high volume) appeal of such a product, no matter how many entertainment experiences and broader optimized applications Apple tries to tempt me with (and no matter how much Apple may drop the price in the future, assuming it even can to a meaningful degree, given bill-of-materials cost and profit-expectation realities). To be clear, this isn’t an Apple-only diss; I’ve expressed the same skepticism in the past about offerings from Oculus-now-Meta and others. And at the root of my pessimism about AR/VR/XR/choose-your-favorite-acronym (or, if you’re Apple, “spatial computing”, whatever that means) headsets may indeed be enduring optimism of a different sort.
Unlike the protagonists of science fiction classics such as William Gibson’s Neuromancer and Virtual Light, Neal Stephenson’s Snow Crash, and Ernest Cline’s Ready Player One, I don’t find the real world to be sufficiently unpleasant that I’m willing to completely disengage from it for long periods of time (and no, the Vision Pro’s EyeSight virtual projected face doesn’t bridge this gap). Scan through any of the Vision Pro reviews published elsewhere and you’ll on-average encounter similar lukewarm-at-best enthusiasm from others. And I can’t help but draw an accurate-or-not analogy to Magic Leap’s 2022 consumer-to-enterprise pivot when I see subsequent Apple press releases touting medical and broader business Vision Pro opportunities.
So is the Vision Pro destined to be yet another Apple failure? Maybe…but definitely not assuredly. Granted, we might have another iPod Hi-Fi on our hands, but keep in mind that the first-generation iPhone and iPad also experienced muted adoption. Yours truly even dismissively called the latter “basically a large-screen iPod touch” on a few early occasions. So let’s wait and see how quickly the company and its application-development partners iterate both the platform’s features and cost before we start publishing headlines and crafting obituaries about its demise.
The M3-based MacBook Air
Fast-forward to March, and Apple unveiled M3 SoC-based variants of the MacBook Air (MBA), following up on the 13” M2-based MBA launched at the 2022 WWDC and the first-time-in-this-size 15” M2 MBA unveiled a year later:
Aside from the Apple Silicon application processor upgrade (first publicly discussed last October), there’s faster Wi-Fi (6E) along with an interesting twist on expanded external-display support; the M3-based models can now simultaneously drive two of ‘em, but only when the “clamshell” is closed (i.e., when the internal display is shut off). But the most interesting twist, at least for this nonvolatile-memory-background techie, is that Apple did a seeming back-step on its flash memory architecture. In the M2 generation, the 256 GByte SSD variant consisted of only a single flash memory chip (presumably single-die, to boot, bad pun intended), which bottlenecked performance due to the resultant inability for multi-access parallelism. To get peak read and (especially evident) write speeds, you needed to upgrade to a 512 GByte or larger SSD.
The M3 generation seemingly doesn’t suffer from the same compromise. A post-launch teardown revealed that (at least for that particular device…since Apple multi-sources its flash memory, one data point shouldn’t necessarily be extrapolated to an all-encompassing conclusion) the 256 GByte SSD subsystem comprised two 128 GByte flash memory chips, with consequent restoration of full performance potential. I’m particularly intrigued by this design decision considering that two 128 GByte flash memories conceivably cost Apple more than one 256 GByte alternative (likely the root cause of the earlier M1-to-M2 move). That said, I also don’t underestimate the formidable negotiation “muscle” of Apple’s procurement department…
Earnings
Last week, we got Apple’s second-fiscal-quarter earnings results. I normally don’t cover these at all, and I won’t dwell long on the topic this time, either. But they reveal Apple’s ever-increasing revenue and profit reliance on its “walled garden” services business (to the ever-increasing dismay of its “partners”, along with various worldwide government entities), given that hardware revenue dropped for all hardware categories save Macs, notably including both iPhone and iPad and in spite of the already-discussed Vision Pro launch. That said, the following corporate positioning seemed to be market-calming:
In the March quarter a year ago, we were able to replenish iPhone channel inventory and fulfill significant pent-up demand from the December quarter COVID-related supply disruptions on the iPhone 14 Pro and 14 Pro Max. We estimate this one-time impact added close to $5 billion to the March quarter revenue last year. If we removed this from last year’s results, our March quarter total company revenue this year would have grown.
The iPad Air
And today we got new iPads and accessories. The iPad Air first:
Reminiscent of the aforementioned MacBook Air family earlier this year, they undergo a SoC migration, this time from the M1 to the M2. They also get a relocated front camera, friendlier (as with 2022’s 10th generation conventional iPad) for landscape-orientation usage. And to the “they” in the previous two sentences, as well as again reminiscent of the aforementioned MacBook Air expansion to both 13” and 15” form factors, the iPad Air now comes in both 11” and 13” versions, the latter historically only offered with the iPad Pro.
The M4 SoC
Like their iPad Air siblings, the newest generation of iPad Pros relocate the front camera to a more landscape orientation-friendly bezel location. But that’s among the least notable enhancements this time around. On the flip side of the coin, perhaps most notable news is that they mark the first-time emergence of Apple’s M4 SoC. I’ll begin with obligatory block diagrams:
Some historical perspective is warranted here. Only six months ago, when Apple rolled out its first three (only?) M3 variants along with inclusive systems, I summarized the to-date situation:
Let’s go back to the M1. Recall that it ended up coming in four different proliferations:
- The entry-level M1
- The M1 Pro, with increased CPU and GPU core counts
- The M1 Max, which kept the CPU core constellation the same but doubled up the graphics subsystem, and
- The M1 Ultra, a two-die “chiplet” merging together two M1 Max chips with requisite doubling of various core counts, the maximum amount of system memory, and the like
But here’s the thing: it took a considerable amount of time—1.5 years—for Apple to roll out the entire M1 family from its A14 Bionic development starting point:
- A14 Bionic (the M1 foundation): September 15, 2020
- M1: November 10, 2020
- M1 Pro and Max: October 18, 2021
- M1 Ultra: March 8, 2022
Now let’s look at the M2 family, starting with its A15 Bionic SoC development foundation:
Nearly two years’ total latency this time: nine months alone from the A15 to the M2.
I don’t yet know for sure, but for a variety of reasons (process lithography foundation, core mix and characteristics, etc.) I strongly suspect that the M3 chips are not based on the A16 SoC, which was released on September 7, 2022. Instead, I’m pretty confident in prognosticating that Apple went straight to the A17 Pro, unveiled just last month (as I write these words), on September 12 of this year, as their development foundation.
Now look at the so-far rollout timeline for the M3 family—I think my reason for focusing on it will then be obvious:
- A17 Pro: September 12, 2023
- M3: October 30, 2023
- M3 Pro and Max: October 30, 2023
- M3 Ultra: TBD
- M3 Extreme (a long-rumored four-Max-die high-end proliferation, which never ended up appearing in either the M1 or M2 generations): TBD (if at all)
Granted, we only have the initial variant of the M4 SoC so far. There’s no guarantee at this point that additional family members won’t have M1-reminiscent sloth-like rollout schedules. But for today, focus only on the initial-member rollout latencies:
- M1 to M2: more than 19 months
- M2 to M3: a bit more than 16 months
- M3 to M4: a bit more than 6 months
Note, too, that Apple indicates that the M4 is built on a “second-generation 3 nm process” (presumably, like its predecessors, from TSMC). Time from another six-months-back quote:
Conceptually, the M3 flavors are reminiscent of their precursors, albeit with newer generations of various cores, along with a 3 nm fabrication process foundation.
As for the M4, here’s my guess: from a CPU core standpoint, especially given the rapid generational development time, the performance and efficiency cores are likely essentially the same as those in the M3, albeit with some minor microarchitecture tweaks to add-and-enhance deep learning-amenable instructions and the like, therefore this press release excerpt:
Both types of cores also feature enhanced, next-generation ML accelerators.
The fact that there are six efficiency cores this time, versus four in the M3, is likely due in no small part to the second-generation 3 nm lithography’s improved transistor packing capabilities along with more optimized die layout efficiencies (any potential remaining M3-to-M4 die size increase might also be cost-counterbalanced by TSMC’s improved 3 nm yields versus last year).
What about the NPU, which Apple brands as the “Neural Engine”? Well, at first glance it’s a significant raw-performance improvement from the one in the M3: 18 TOPs (trillion operations per second) vs 38 TOPs. But here comes another six-month back quote about the M3:
The M3’s 16-core neural engine (i.e., deep learning inference processing) subsystem is faster than it was in the previous generation. All well and good. But during the presentation, Apple claimed that it was capable of 18 TOPS peak performance. Up to now I’d been assuming, as you know from the reading you’ve already done here, that the M3 was a relatively straight-line derivation of the A17 Pro SoC architecture. But Apple claimed back in September that the A17 Pro’s neural engine ran at 35 TOPS. Waaa?
I see one (or multiple-in-combination) of (at least) three possibilities to explain this discrepancy:
- The M3’s neural engine is an older or more generally simpler design than the one in the A17 Pro
- The M3’s neural engine is under-clocked compared to the one in the A17 Pro
- The M3’s neural engine’s performance was measured using a different data set (INT16 vs INT8, for example, or FLOAT vs INT) than what was used to benchmark the A17 Pro
My bet remains that the first possibility of the three listed was the dominant if not sole reason for the M3 NPU’s performance downgrade versus that in the A17 Pro. And I’ll also bet that the M4 NPU is essentially the same as the one in the A17 Pro, perhaps again with some minor architecture tweaks (or maybe just a slight clock boost!). So then is the M4 just a tweaked A17 Pro built on a tweaked 3 nm process? Not exactly. Although the GPU architecture also seems to be akin to, if not identical to, the one in the A17 Pro (six-core implementation) and M3 (10-core matching count), the display controller has more tangibly evolved this time, likely in no small part for the display enhancements which I’ll touch on next. Here’s the summary graphic:
More on the iPad Pro
Turning attention to the M4-based iPads themselves, the most significant thing here is that they’re M4-based iPads. This marks the first time that a new Apple Silicon generation has shown up in something other than an Apple computer (notably skipping the M3-based iPad Pro iteration in the process, as well), and I don’t think it’s just a random coincidence. Apple’s clearly, to me, putting a firm stake in the ground as to the corporate importance of its comparatively proprietary (versus the burgeoning array of Arm-based Windows computers) tablet product line, both in an absolute sense and versus computers (Apple’s own and others). A famous Steve Jobs quote comes to my mind at this point:
If you don’t cannibalize yourself someone else will.
The other notable iPad Pro enhancement this time around is the belated but still significant display migration to OLED technology, which I forecasted last August. Unsurprisingly, thanks to the resultant elimination of a dedicated backlight (an OLED attribute I noted way back in 2010 and revisited in 2019) the tablets are now significantly thinner as a result, in spite of the fact that they’re constructed in a fairly unique dual-layer brightness-boosting “sandwich” (harking back to my earlier display controller enhancements comments; note that a separate simultaneous external tethered display is still also supported). And reflective of the tablets’ high-end classification, Apple has rolled out corresponding “Pro” versions of its Magic Keyboard (adding a dedicated function-key row, along with a haptic feedback-enhanced larger trackpad):
And Pencil, adding “squeeze” support, haptic feedback of its own, and other enhancements:
Other notable inter- and intra-generational tweaks:
- No more mmWave 5G support.
- No more ultra-wide rear camera, either.
- Physical SIM slots? Gone, too.
- Ten-core CPU M4 SoCs are unique to the 1 TByte and 2 TByte iPad Pro variants; lower-capacity mass storage models get only 9 CPU cores (one less performance core, to be precise, although corresponding GPU core counts are interestingly per-product-variant unchanged this time). They’re also allocated only half the RAM of their bigger-SSD brethren: 8 GBytes vs 16 GBytes.
- 1 and 2 TByte iPads are also the only ones offered a nano-texture glass option.
Given that Apple did no iPad family updates at all last year, this is an encouraging start to 2024. That said, the base 10th-generation iPad is still the same as when originally unveiled in October 2022, although it did get a price shave today (and its 9th-generation precursor is no longer in production, either). And the 6th-generation iPad mini introduced in September 2021 is still the latest-and-greatest, too. I’m admittedly more than a bit surprised and pleased that my unit purchased gently used off eBay last summer is still state-of-the-art!
iPad software holdbacks
And as for Apple’s ongoing push to make the iPad, and the iPad Pro specifically, a credible alternative to a full-blown computer? It’s a topic I first broached at length back in September 2018, and to at least some degree the situation hasn’t tangibly changed since then. Tablet hardware isn’t fundamentally what’s holding the concept back from becoming a meaningful reality, but then again, I’d argue that it never was the dominant shortcoming. It was, and largely remains, software; both the operating system and the applications that run on it. And I admittedly felt validated in my opinion here when I perused The Verge’s post-launch event livestream archive and saw it echoed there, too.
Sure, Apple just added some nice enhancements to its high-end multimedia-creation and editing tablet apps (along with their MacOS versions, I might add) but how many folks are really interested in editing multiple ProRes streams without proxies on a computer nowadays, far from on an iPad? What about tangible improvements for the masses? Sure, you can use a mouse with an iPad now, but multitasking attempts still, in a word, suck. And iPadOS still doesn’t even support the basics, such as multi-user support. Then again, there’s always this year’s WWDC, taking place mid-next month, which I will of course once again be covering for EDN and y’all. Hope springs eternal, I guess. Until then, let me know your thoughts in the comments.
p.s…I realized just before pressing “send to Aalyia” that I hadn’t closed the loop on my earlier “building block mass-production availability delays” tease. My suspicion is that originally the new iPads were supposed to be unveiled alongside the new MacBook Airs back in March, in full virtual-event form. But in the spirit of “where there’s smoke, there’s fire”, longstanding rumors about OLED display volume production delays, I’m also guessing (and/or second-generation 3 nm process volume production delays), pushed the iPads to today.
—Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.
Related Content
