As a pseudo-tech snob watching the most consequential Apple pro-focused (Mac) event since, I dunno, maybe the Intel transition?!, I've been keenly interested in what kind of technology base Apple would use when the new pair of MacBook Pros finally launched.

I don't think it was unreasonable of me to expect something newer (and which just shipped!) given the passage of time. But in the end, I'm left feeling more than a little silly, because scaling up the M1's architecture,

as Apple Silicon all-star and Hardware Technologies SVP Johny Srouji specifically said at least twice during Monday's presentation,

is a heck of a lot more involved than just "gluing" CPU, GPU and NPU cores together.

Time to get right into why the M1 Pro/Max SoC series seem

• thoroughly modern, even if "several months late"

• maybe a little bit too serious (I'll explain what I mean soon enough)

• relentlessly focused (you've heard this phrase before, but this week, Apple provided even more evidence)

An Unapologetically 2021 SoC, Despite Possible (and Quite Understandable) Pandemic-Related Delays

How can I say with such confidence that the core technology of the M1 Pro and Max is "Late 2020, M1-derived"? Even before the inevitable early Geekbench 5 results roll in, there are at least two clear signs:

First, the GPU cores are in units of seven and eight, a clear parallel to the binned configurations in the M1 MacBook Air and more recently introduced M1 iMac. Not only that, Apple specifically showed off the sheer parallelism of the M1 GPU architecture in two slides:

5.2 TFLOPS of performance from double the GPU cores of M1...which has a claimed 2.6 TFLOPS of performance.

Double the 16-core GPU complex from the top-binned M1 Pro to 32 cores in the top-binned M1 Max, and...

...you get a pretty impressive (especially for a laptop!) 10.4 TFLOPS, precisely quadruple the rated performance of the M1's GPU (as Srouji pointed out moments later)

Second, the onboard M1 Pro/Max Neural Engine handles 11 trillion operations/second - same as the A14 Bionic and M1. Here's part of the Apple Event slide, noting the Neural Engine's performance rating in some amusingly small type (repeated for the M1 Max):

Since both the GPU and Neural Engine don't resemble the higher-performing versions found on the A15 Bionic, that means the CPU cores, too, are almost certainly "M1"-derived.

There's much more to it than that, however. As Srouji said in no uncertain terms for the "entry-level" pro-grade M-chip: "we...scal[ed] up M1's...architecture to create a far more powerful chip with M1 Pro" (around 17:05 into the presentation)

And it shows, with advertised new features never seen in any Mac to date:

• On-SoC ProRes encode/decode engine(s) (similar to iPhone 13 Pro, probably a lot more industrial-strength)

• Faster LPDDR5 unified RAM and 200GB/s or 400 GB/s memory bandwidth, far outpacing M1's ~68.2GB/srating as calculated by Andrei Frumusamu of AnandTech

• an entirely different set of Apple Silicon perf curves for CPU and GPU, as I'll recap later.

Not a Single Reference to Mac Gaming (Even Though the MBP 14.2 and 16.2 GPUs Have the Capability)

New GPU SKUs with ~4.5,5.2, 7.8 and 10.4 TFLOPS ratings based on GPU core counts. All of which are theoretically at respectable-to-actually-quite-fast discrete-laptop-GPU level. For example, the top-end 32-core M1 Max GPU has some TFLOPS and texture fill rate (gigatexels/sec) similarities with the 135W max power-rated AMD Radeon RX 6700M, except for potential direct access to far more (albeit non-GDDR6) unified memory, plus 25% more memory bandwidth than the discrete 6700M GPU.

So the new MacBook Pros should be credible gaming machines. Yet, there was zero mention of M1 Pro/Max's gaming potential throughout the entire Mac portion of the presentation. The lucky devs chosen to preview the M1 Pro/Max systems this time around represented DaVinci Resolve, Adobe Premiere Pro, Octane X, Cinema 4D, and yes, Unity Editor...but of course, in the context of creating game assets.

Is this a somewhat worrying sign to a continually shrinking 😭 set of increasingly discouraged Mac gamers that Apple just doesn't care that much about AAA-type titles on Mac?

Or was Apple simply trying to direct their product marketing efforts towards that likely-majority segment of Apple's performance MacBook Pro demographic - a group that could generally care less about games, rather different than the iPhone 13 Pro mobile gaming demographic?

Maybe that yet-to-be-replaced iMac 5K, once on Apple Silicon, will give us more clues. And I get it. To show off the ability to play AAA games with ease, it helps to have some modern AAA games optimized for M-chips, and Baldur's Gate 3 was already shown off for the M1 Macs last November. That top-tier Mac-native game shelf will remain sparse for a while yet.

Laser-Focused...on Delivering the Best Damn Balanced Performance Laptops in the Industry

Apple is likely to have shown a considerable lead in mobile CPU multiprocessing, which would crown M1 Pro/Max as fastest in that key compute area amongst all laptops.

However, Apple appears to be very comfortable with not being the unquestioned leader in mobile GPU performance or perhaps even single-threaded mobile CPU performance this round.

Why?

The answer, it seems, is found within the mission of the new MBP 14.2 and MBP 16.2, which itself is a reflection of what certainly looks like the extreme discipline of the Apple Silicon teams. Srouji and company want M1 Pro and M1 Max to be the hands-down leaders in performance-per-watt, and it looks like they're strictly following these performance directives:

• No power-hungry turbo clocks for single-core (not yet or not for MacBook Pros, anyway).

• 30W maximum as a general rule for CPU multiprocessing.

• Around 30W max as a general rule for the M1 Pro GPU.

• About 60W max as a general rule for the M1 Max GPU.

The likely proof is in the claimed performance curves. First, the 16-core-bin M1 Pro GPU comparison:

Here, M1 Pro is comfortable being slightly faster than what appears to be an onboard GeForce RTX 3080 - plenty powerful in desktop form, but apparently thermally constrained to a "mere" 105W within a 0.67"-depth laptop. [CORRECTION: The GPU in question looks to be a GeForce RTX 3050 Ti, a considerably more modest mobile GPU with an apparently dodgy perf curve north of 70-80W. My apologies for mixing it up with the Razer Blade Advanced 15's RTX 3080, which per Apple's claims is a close competitor of the M1 Max.] It seems like M1 Pro could gain appreciable performance by extending to 40 or even 50 watts, assuming the GPU complex was rated to handle that extra power. But that is simply not the Apple Silicon Way. In the 100W-hr absolute battery limit world of laptops, Apple prefers to let the extreme difference in perf/W do the talking.

Apple's very different view of performance bragging rights is arguably even better exhibited when M1 Max competes with a beast of a gaming laptop equipped with a GPU chewing through over 160W of power (yes, more than an entire 143W M1 iMac power supply):

Does it look like Apple Chipworks could give M1 Max another 20-30 watts to get over that 400 score on the relative performance scale, allowing it to claim M1 Max as possibly the fastest laptop GPU in the world? It does.

Would Srouji and his team actually allow an Apple Silicon laptop GPU to spike to 80-90W total, for any reason? I seriously doubt it, because "giving up" perhaps 6-7% of performance in exchange for saving around 100 watts of power versus the discrete GPU competition is very much worth the tradeoff. (Plus it makes for a great product marketing compare.)

To wrap up, did Apple Chipworks kinda, sorta, leave something on the table? While they had their reasons, and damn good ones...yes, they obviously did. Just look at the M1 Pro/Max multi-core CPU perf curve claim:

It's clear from all of these truncated performance curves that there's still higher and still faster out there for the taking. But for the Apple Silicon team, it would mean crossing an unseen (to us) bright red line of unacceptable diminishing returns.

Somehow, I doubt any significant number of MBP14/16 users (who're using these machines as Apple customers, not tech reviewers) are ever going to care. For Mac, Apple has recommitted to performance-per-watt over all else, reprioritizing function and customer-favorite ports over some endless drive towards thinner and lighter. In the process it's seemingly delivered a duo of laptops that, while not necessarily dominant in every specification, are nonetheless ultra-versatile, high-performance/super-high-efficiency traditional-type laptops (with some awfully nice variable-refresh rate displays) which will send even more Apple Silicon shockwaves throughout the personal computing industry.

Not bad for an M1 variant.