Painted Into an iPad-Paradigm Corner?
Once, there was Apple A4.
True, some Apple Silicon watchers see A4 as something more like a chip “co-branded” with then-foundry partner and at-the-time-increasingly-frenemy Samsung, rather than something full-custom like Apple does now and has since at least A6. But there’s no question about what A4 enabled (aside from the first iPad). It powered a jaw-dropping 326 ppi display, great performance, with really-quite-decent battery life, wrapped up in a “glass and steel sandwich” iPhone 4 design many see as the pinnacle of smartphones (at least, those with now-considered-tiny 3.5″ displays).
But barely one generation later, Apple needed to spin off a variant of its dual-core A5 (used in iPhone 4S and iPad 2) in order to enable a high-resolution display for its new-and-skyrocketing iPad business. And so the first “AX”-chip, the A5X, arrived with the additional GPU power needed to drive 4x the pixels of the 1024×768 iPads before it.
As Johny Srouji so helpfully reminded us last year, the mostly-iPad-specific, higher-powered A-chips have seen six (main) generations: A5X, A6X, A8X, A9X, A10X, and A12X.[FN1]
When Apple was finally ready to release Apple Silicon Macs, it used the same technology base as the A14 Bionic. While Apple has been careful to clarify that M1 is not merely some higher-performance variant of A14, the (ahem) core comparisons are simply too obvious to ignore:
• 4 performance CPU cores vs. 2 (with similar Geekbench 5 results)
• 7 or 8 GPU cores vs. 4
• the same Apple Neural Engine (the 3rd-gen one operating at 11T ops/sec)
• a unified memory architecture that looks, at least superficially, exactly like the basic SoC-DRAM layout of the A12X (Apple SoC on the left, two side-by-side DRAM modules immediately to the right)
I get why Apple is justifiably proud of the M1, and why it doesn’t want to give the impression that it’s an “A14X” thrown in a Mac. But…from the looks of things, that’s pretty close to what happened, and as it so happened, the M1 found its way from Macs…right into the 2021 iPad Pro lineup. At least to me, though, it’s really just fun tech trivia, and absolutely nothing to “apologize” for.
The fact is, Apple has worked so diligently on perf/W on iPhones and for so long, it later found this basic A-chip architecture scaled well enough to power its entire consumer-level Mac lineup at a basically-prosumer overall level of performance (GPU aside). It does this by combining ridiculously powerful CPU performance (again, entry-level chips at Core i5-ish system price points) and very fast I/O + storage with the power-sipping benefits that a finely-tuned, highly-optimized state-of-the-art 5nm lithography process makes possible. Sure, the A14 platform had hard limits on RAM, SSD, and I/O bandwidth beyond two Thunderbolt 3 ports and two standard USB-C ports. But these are really quite minor tradeoffs for desktops and laptops that have no business being this fast at $699/$999 starting prices, and I have little doubt Apple’s hardware engineering teams will look to improve on DRAM and I/O limitations over the next several years.
Still, this does present a bit of a marketing problem. Consumers are used to A15 this and A12X that, with the “X” standing for “more performance”. So now that we’re anticipating the next tier up from the M1 chip for Macs (MBP16, bigger iMacs, possibly a MBP14 and a Mac mini Pro), does it logically follow that Apple calls this new Apple Silicon Mac M-chip an “M1X”? Around a year after the M1 already debuted, with rumors of an M2 just around the corner?
An Elegant Marketing Solution, At Least
Why is “M1X” a “problem”? Putting aside the nigh-imperceptible impact on market uptake (because prosumer Mac buyers just want fast, and Apple is primed to deliver), it’s really a matter of timing. The future of prosumer Macs and MacBook Pros have taken so long to get here, Apple’s actually launched what looks to be a TSMC N5P-based A15 in the interim. In other words, an SoC based on leading-edge second-run (optimized) 5-nanometer lithography, throughly 2021 tech for lack of a better description. M1, great as it is, is 2020 tech.
So there’s Apple not wanting to tie the “halo Macs” ready to be unleashed on the world to “last year’s chip”.
There’s also the matter of confusing buyers, especially if Apple does not announce its next-generation entry-level Macs at the same time as the prosumer Macs. The rumor community has had a tougher go of it lately given their cat-and-mouse game being escalated in a major way by a leak-weary Apple, but their best guess is that Apple’s only announcing prosumer Macs at its next Apple Event involving…well…new Macs.
M1, for better or for worse, means entry-level Macs through MacBook Pro 13″ and iMac 24″. So M2 logically means the next M-chip update for this same class of Macs.
Apple could announce M2X first, but Apple’s never launched an “X variant” chip in advance of the normal version. With a big assist from Wikipedia:
• Apple A5 (iPad 2, Mar. 2011) >> A5X (iPad 3, Mar. 2012)
• A6 (iPhone 5, Sep. 2012) >> A6X (iPad 4, Nov. 2012)
• A8 (the infamous iPhone 6, Sep. 2014) >> A8X (iPad Air 2, Oct. 2014)
• A9 and A9X (simultaneous announcement for iPhone 6s and iPad Pro, Sep. 2015)
• A10 (iPhone 7, Sep. 2016) >> A10X (iPad Pro 2, Jun. 2017)
• A12 (iPhones XR, XS and XS Max, Sep. 2018) >> A12X (iPad Pro 3, Oct. 2018)
If Apple wants to name its chips in the same general way as A-chips, but also in a way that can easily associate a chip name with a Mac family and tech generation, it could swap a letter with a number and do this:
MX1.
“M” for, of course, the Mac version of Apple Silicon. “X” for the same “higher performance” we know from many iPads. And “1” for the first Apple Silicon chip family specifically geared towards these “X-tier”, prosumer-focused higher-performance Macs.
Somewhat similar to how automakers have a “First Edition” of a new generation of a performance variant, Apple can communicate with two letters and a number about how it’s (planning to) put the rest of the higher-performance PC world on notice, redefining what best-in-class means for compute performance in everything short of (1) high-core-count workstations/servers and (2) kilowatt-class gaming/crypto-mining 🥴 PCs with triple-fan cooling (just for the GPU!), with neither HPC category really giving all that much of a damn about “power efficiency” in any normal consumer sense.[FN2]
“MX1” also has the side benefit of allowing Apple, if needed, to release prosumer-level Mac hardware updates without worrying about the consumer-side product cadence. While I suspect a large part of the Apple Silicon shift was to allow Apple to upgrade the SoCs of its higher-volume Macs as frequently as Apple updated its A-chips…who knows? Perhaps there will be a point in the future where Apple needs up to two years to launch a new Mac SoC…certainly something that sounds possible on its highest-performing Mac, the lowest-volume Mac Pro that I have to think Apple will not give up on before completing the Apple Silicon transition.
Are there potential problems with MX1? Sure – given the almost one year that’s passed since MacBook Air, MBP13 and Mac mini were given the M1, the “1” still leads to associations with M1, even though the underlying chip architecture might well resemble what many Apple Silicon watchers would consider an M2. And the “MX” thing may come with “IP baggage”, whether we’re talking Mazda (Miata, also known as MX-5) or even brand-new bitter former-GPU-partner-turned-competitor NVIDIA (which still uses its MX branding to this day). Then again, Logitech creates “MX” brand keyboards and (computer) mice, and I haven’t heard of either Mazda or NVIDIA trying to sue it out of existence. And Germany-based computer peripherals company Cherry produces its own “MX” line of keyboards while apparently not being sued by Logitech. So…maybe “MX” isn’t much of an IP gauntlet after all depending on how nicely Apple asks, or how well they can thread their branding needle.
Commonality with M1, Short-Term and Longer-Term?
I’ll freely admit to a limited imagination and total non-expertise here, but for now, it seems likely that Apple can scale up the “building blocks” of its first Apple Silicon Mac SoC (CPU and GPU cores), use the modernized A15-platform versions, add some RAM and expand I/O, and boom, it has everything it needs to both thrill a large segment of the Mac prosumer market and send a chill through the PC competition.
But will it always be that way, and should it? Perhaps pro workflows will emerge that require dedicated hardware (see Apple’s Mac Pro-exclusive Afterburner add-on card), or maybe even a fundamentally more sophisticated hardware architecture + motherboard layout that the lower-power consumer Mac market won’t benefit from. Just to give one (very important) example, in the world of system DRAM, there’s LPDDR4X which is plenty for consumers, but HBM2, despite the higher cost and theoretical latency, delivers peak memory bandwidth that looks like a friggin’ superhighway (Samsung claims a wild 1TB/s with its “Flarebolt” technology) compared to LPDDR4X’s quieter, narrower two-way country road (Apple claims around 68GB/s in unified memory bandwidth).
Sure, until an Apple Silicon Mac Pro comes along, an MBP16 CPU core and GPU core by themselves may look quite a bit like a MacBook Air CPU core and GPU core. But that could well change over time, and if so, it won’t be by accident.
Coming Up: Possible Ways to Tell Whether the Please-Arrive-Before-2021-Ends Next-Generation Prosumer Macs are Running “This Year’s” or “Last Year’s” Tech
Since this topic doesn’t quite fit in the context of this post, I’ll write up a bonus post on how we can “tell”, right off the bat, what type of chip Apple’s bringing to what I assume will be its entire prosumer (call it “premium performance just short of Mac Pro”) Mac lineup. And maybe I’ll finally get around to making some just-for-fun GB5 score predictions as well.
—————
Footnotes:
[FN1] (A12Z, sort of a “higher-binned” A12X, added one GPU core over A12X, for a total of eight)
[FN2] In pure unit sales volume terms, the next performance M-chip’s relevant competition will be laptops. And gaming laptops (a category I don’t see MBP16 competing with…directly…right away) will likely be able to stuff in even more sheer GPU firepower than even a top-spec first-generation M-chip MBP16, since Apple is about efficiency-minded, balanced systems with as-respectable-as-possible battery life. Also, gaming laptops have their own significant tradeoffs. One well-known brand, Razer, has a laptop line with 14″ 4MP displays, some running an imposingly performant NVIDIA GeForce RTX 3080, but requiring a 230W adapter (yes, as in one and a half entire iMac 24 power supplies’ worth) to power it all…because c’mon, the on-board ~60W-hr battery is not for those intense e-sports sessions.
The AAPL Tree 