Apple Silicon M2 — The Fine Print You Might Have Missed

We can all agree that WWDC 2022 was great, but the devil’s in the details…

WWDC 2022 came and went. If you ask me, it was the greatest of them all. Incredibly dense and packed with information. I remember at one point I told my buddy Andrew Gribben, as we were watching it together, to shut up because I didn’t want to miss the details. Regardless, we still missed, we missed plenty of them. Even if you paid all the attention in the world, you were bound to miss something, so it’s no surprise that I felt I needed to look a lot deeper into the M2 architecture announced, and see what’s what, and share some of my findings with y’all.

I think that while many — including myself — see the M2 as just an iteration, in some ways we are all being unfair on Apple. Yes, in 2020 Apple dropped the entire world’s jaws and released something admittedly revolutionary, but let’s not forget how many years have gone into reaching that revolution. Over a decade. Expecting a technological revolution or seismic shift every couple of years is not sustainable. Not from Apple’s perspective, not from the user’s perspective. I think nerds like us need to stop expecting miracles, even from trillion-dollar companies. Physics, economics, geography, politics, and everything else still applies to Apple as well. Modern miracles do take a lot of time.

I think we all have to remind ourselves that M2, while an iteration, it’s an iteration on something that changed the PC industry overnight. Iterating on excellence is no small feat!

I have seen countless reviewers, yes, even some of the big ones out there, being all “meh” about the 18% increase in performance. Allow me to let you in on a little secret. That’s actually very much within industry standards, and it’s on the better end of it, too. Say a process used to take 5 hours, and now it takes just a little above 4. In my book, that’s a pretty big improvement, and anyone who ever had to wait for stuff to get processed will agree with me.

Interestingly enough, what gets a few more folks these days excited is GPU performance. That 35% increase is indeed pretty much the double of the CPU increase, but this will likely be more useful to fewer people than that snappier M2 CPU. Having said that, 35% is nothing to sneeze at.

But that’s not the whole story, is it? After all, this is an SOC and none of its submodules ever run in complete isolation. When you add together the CPU, GPU, the 40% faster neural engine and 50% more memory bandwidth, the numbers all together, which will often be the case when using a machine in the real-world, give us a much more exciting picture. How’s so, you might ask?

With the M2, on average you get a 25% faster machine for free. Free, because it doesn’t cost you a single dime more to run than the M1. This makes M2, and the machines running it, about 25% more cost-efficient.

If saving cost on electricity is your main concern, and honestly, I can’t see why it wouldn’t in an increasingly expensive energy situation across the world, the M2 makes any machine a viable option. Maybe not for those already owning an M1-based machine, but anyone on Intel should make the move, make it now, and preferably to the M2.

Now, interestingly enough, that aforementioned 25% cost-efficiency also just so happens to match up with the SOC being built on a 2nd generation 5-nm technology. This is important to remember. Most folks will just stop reading at 5-nm and move on, assuming it’s the same old thing as two years ago. They’d be very wrong. That 2nd generation aspect is what enabled the M2 to be an M2. If we were to wait for the 3-nm process, we’d have no new Apple hardware topics to talk about right now. We’d all be sad and frustrated like Jon Prosser and Linus, but they do that to themselves, so whatever. 😆

Small-print is small-print for a reason…

But of course, the so-called small-print is only small-print because it’s meant to avoid highlighting aspects that might be seen as shortcomings. One of those is the continued support for only one external monitor. I wrote about this before, I genuinely don’t believe it’s a problem for 95% of users, but it is nevertheless a limitation one needs to keep in mind.

But back to those performance numbers. If you read carefully Apple’s own press release about the M2, and read that literal small-print, you will notice one very interesting detail — the amount of RAM used for testing. This is both good and bad news. Certainly good for those wanting to buy an M2 machine with 16 GB of memory, as they will already know exactly what to expect in terms of performance. It’s also excellent news for those wanting to go even higher, to 24 GB of memory. They will most definitely get more performance out of their machines, though I wrote an entire article why one should not opt for a 24 GB M2 machine.

It is however not so great news for those opting for the very base model of only 8 GB of memory. That’s where Apple is being a bit suspiciously silent. Don’t get me wrong, I am not saying the difference is major, but it will be less, and even more so at sustained loads on the M2 Air where there are no fans. You’ll also have to account for SWAP memory being used a lot, which, even with a fast SSD, will slow the machine down. Will it be as slow as an Intel? God, no! The M2 in its base configuration will still be a worthy machine, but that average total 25% performance and power-efficiency gain, is not something I would vouch for any more.

Let me end on a positive note, though — single-core performance. I bet most folks didn’t think much about this, but here’s the fascinating bit. The M2 can actually be expected to be faster than not just the M1 but also the M1 Pro and M1 Max in single-core performance, because learning from the M1’s architecture, Pro, Max and Ultra, all achieved virtually the same scores in Geekbench. Where the true difference was visible, was multi-core. That makes that initial 18% CPU gain even more attractive, because more than half the apps we run daily, still run on a single core, and don’t make use of multi-core architectures.

M2 gives us an insight into what Apple Silicon iterations will look like going forward. Once everyone understands you only get miracles once a decade, it will be appreciated for what it is — a better SOC than last year’s.

Attila Vago — Software Engineer improving the world one line of code at a time. Cool nerd since forever, writer of codes and blogs. Web accessibility advocate, Lego fan, vinyl record collector. Loves craft beer!