I’ve been thinking a lot about the Apple M1 system on a chip (SoC) since its announcement a couple of weeks ago. More than any normal person should think about a SoC. As a hobbyist PC builder who has custom built many PC’s over the last 23 years and has kept a pulse on the latest CPU technologies and advances, I can’t recall a time where a new CPU or technology has made such an enormous leap over everything else on the market. Not the switch from 386 to 486, not the advancement from Pentium to Pentium II/III/Dual Core/i-series/etc., not the transition from AMD K5 to K6/Athlon/XP/64/x2/etc. The CPU performance leap that Apple has achieved with the M1 is monumental. Not only does it have faster single core and multi-core performance than all current mid-range CPU’s on the market, it’s trading punches with AMD and Intel’s higher-end processors. And it’s doing all of this at a fraction of the voltage (wattage) and being passively cooled (no heatsinks or fans). Apple’s M1 emulates x86 performance better and more efficiently than true x86 processors! All of this with a first generation chip! It’s impressive what they’ve accomplished. This is another 2007 iPhone announcement and product release-type moment and it’s hardly being noticed.
Just wait until someone figures out a way to emulate or virtual machine Windows 10 with the M1 and shows that it performs on-par or better than Windows 10 running natively on any AMD or Intel-based machine. Then people might start to notice…
An advantage that the M1 SoC has is good integrated graphics. Current benchmarks of the M1’s GPU put it on-par with a NVIDIA GTX 1050ti, a 2+ year-old graphics card by today’s standards. Today, that performance is more than adequate for desktop applications (web, Office, etc.), most photo and video editing, and for some 1080p gaming. Intel Iris and AMD Vega are getting much better than the old days of integrated graphics, but neither are offering a SoC. Given Apple’s year-over-year incremental improvements to its A-series processors in its phones and tablets, I’d wager a bet that by the 2nd or 3rd generation of the M-series chip that the graphics performance will catch-up to whatever is the standard at the time of its release.
Another consideration is that the M1 chip is what Apple is showing us is powering their entry-level products: the Mac Mini, the MacBook Air, and the MacBook Pro (13″). What they haven’t mentioned is what’s in store for its iMac, iMac Pro, Mac Pro, and MacBook Pro (16″) lineups. These machines have typically been powered by Intel’s mid to higher-end CPU’s (more cores and threads, higher voltage), have been spec’d out with higher amounts of RAM, and with higher-end GPU’s. I’m sure whatever Apple has planned for these machines is going to be a leap beyond what they have shown us with the M1.
This tweet from DHH,
I pray that Intel, AMD, and Qualcomm is letting the M1 give them ideas, take them in new directions. Because this level of sorcery is too damn powerful to be held by a single company. Especially a monopolizing conglomerate like Apple. But fucking kudos to those chip wizards 👏
AMD, Qualcomm, and certainly not Intel have a standalone CPU in their pipeline to compete with the performance of Apple’s M1. None have a SoC that competes with the package that Apple is offering with the M1. I do hope that they are paying attention and can innovate and compete. I can still see a future where hobbyists custom build their own PC’s, but at the price, performance, packaging, and value that Apple is proposing, it may upend that hobby and market.
[update 11/27/2020] Developer successfully virtualizes Windows for ARM on Apple M1
[updated 11/30/2020] I edited my original post after reading Erik Engheim’s article “Why is Apple’s M1 Chip So Fast?” because where I meant “CPU” I actually meant “SoC.”
[updated 12/3/2020] Benchmarks show that the ARM version of Windows 10 running on the Apple M1 destroys the ARM version of Windows 10 running on the Microsoft Surface Pro X. From the article:
Windows on an M1 got a single-core score of 1,288 and multi-core score of 5,685 whereas the Surface Pro X’s scores were roughly 800 and 3,000 in those respective benchmarks.
That’s 1.5x the single-core performance and almost 2x the multi-core performance. The Ryzen 3 system I built just three months ago is starting to seem like a relic from year’s past.