The new top-of-the-line Apple Silicon chip comes with one surprise and potentially very significant performance advantages in media-intensive and software development applications.
The most interesting segment of Apple’s March 8 Event dealt with the M1 Ultra chips — plural intended and explained below. Today, we’ll focus on what this latest member of the M1 family means, on what it portends for future Apple products.
A brief history: the M1 chip was announced at Apple’s June 2020 WWDC (it’s yearly developers’ conference) with an overall theme and several promises. The new member of the Apple Silicon family (the Wikipedia article provides valuable perspective) set a tone that was to become common to all future announcements: More Power, Less Watts. In practical terms, your M1-equipped laptop would be faster and its battery would last longer — when compared to Intel-powered devices, of course. This came with one shorter term promise, a MacBook Air and a MacBook Pro to ship before year-end, and a longer one: all Macs would be converted to Apple Silicon within two years.
About a year later two new M1 processors were announced: M1 Pro and M1 Max, larger chips with, in the ascending order of their names, more of everything: memory, CPU, GPU and Neural Engines cores. These late 2021 chips powered two MacBook Pro laptops in 14”and 16” diagonal screen dimensions. Real “pros” I spoke with, who develop software, rave about the speed and battery life of the new devices, the 16” version in particular.
So, when rumors started of a new Mac to be announced early March 2022, there was speculation that Apple would be announcing a newer M2, more powerful member of the Apple Silicon family, perhaps based on a newer TSMC 3 nm process — as opposed to the 5nm version used for M1 devices. (Here 5 nm and 3 nm refer to the size in nanometers, billionths of meter, of circuit elements used to build chips. In general terms, smaller elements mean denser, more powerful chips.)
Surprise, Apple announced not an M2 device but a new member of the M1 family, the M1 Ultra. Further surprise, the M1 Ultra actually consists of two M1 Max chips joined thanks to a silicon interconnect feature that had been all along part of the device. (Fans of Johny Srouji. yours truly included, will delight in his exposé of the M1 Ultra starting at 26 minutes into the show. Here Johny proves to be as excellent a pitchman as he is as the senior head of Apple Silicon developments). In essence, Srouji says, in the M1 Ultra, the two M1 chips are joined without paying much of the penalty associated with connecting two processors together: power consumption, bandwidth and memory management. For example, the UltraFusion interconnect, as it is christened, manages more that 10,000 signals with a bandwidth that stands at 2.6 terabytes per second. Memory stays unified (no changes to applications) up to 128 gigabytes and the combined device holds 114 billion transistors, the largest amount ever in a desktop chip. In “fusing” the two M1 Max chips, Srouji and his team come up with a never-before 64-core GPU The M1 Ultra feature set is summarized here:
This allows Anandtech, a leading technology site to conclude its first look at the M1 Ultra thusly [as always, edits and emphasis mine]:
“The net result is that Apple has announced a SoC that has no peer in the industry across multiple levels. Going multi-die/multi-chip in a workstation is a tried and true strategy for CPUs, but to do so with GPUs will potentially put Apple on a level all of their own. If their transparent multi-GPU technology works as well as the company claims, then Apple is going to be even farther ahead of their competitors both in performance and in developing the cutting-edge technologies needed to build such a chip.
Great. But there are a few buts.
First, benchmarks will reveal that, for a single thread, a single sequence of operations, the M1 Ultra isn’t faster than an entry-level M1 chip. This is because the the clock speed associated with the 5nm process common to all M1 chip hasn’t changed for the M1 Ultra. The newer chip will particularly shine in multithreaded applications generally associated with media development (audio, video, animation…) and some software development. All of which constitute a juicy and traditional enough market for Apple whose control of its macOS system software helps maximize multithreading performance.
Second, the recourse to two M1 Max chips fused into a M1 Ultra means TSMC’s 5 nm process has reached its upper limit. It also means TSMC’s 3 nm process isn’t ready, probably not shipping until late 2022. Apple, by virtue of their tight partnership with TSMC has known about and taken precautions against the 3 nm schedule, hence the initially undisclosed M1 Max UltraFusion design wrinkle, likely an early 2021 decision.
Third, the new M1 silicon needs a new home. Roughly speaking, it could dissipate as much as twice the heat of a M1 Max chip found in a late 2021 MacBook Pro. This leads to a new desktop, the Mac Studio, the name a clear reference to applications in which it is expected to shine. Built on a single aluminum extrusion, with the same footprint as the Mac mini, 7.7 inches square and almost 3 times as high at 3.7 inches. The Mac Studio’s additional height makes room for a bigger fan needed by the M1 Ultra’s larger power dissipation. (In passing, we’ll note there is Mac Studio version sporting a single M1 Max chip that happens to weigh 2 pounds less, most likely the result of a smaller power supply.)
Fourth, to go with Mac Studio intended uses, a 27” Studio 5K display costs $1599 in its basic version — nano-textured glass being $300 extra. While the tilt-adjustable stand is included in the $1599 price, a tilt- and height-adjustable stand costs an extra $400… This is perhaps a challenge and an incentive to go to sites such as ergotron.com that feature a range of monitor arms solutions, some with cable management features. Still, the new Studio monitor is welcome alternatives to the 32” Pro Display XDR “starting” at $4999 and demanding an extra $1,000 for the nano-textured glass — plus $999 for a Pro Stand. It’s good to see Apple back into the somewhat affordable monitor game, especially when powered by an A13 CPU found in the Fall 2019 iPhone 11 driving a 12 megapixels Center Stage Camera and a 12-speaker sound system, and offering up to 96 watts of power to charge one’s laptop. While a capable 27” 5K display made by LG is offered at about $300 less, Apple surely expects the Studio’s extra features and aesthetics will help sell it.
Fifth, the Intel-based 27” iMac (on which I write this column) is no more and unofficially given little chance of being born again with an M1 variant.The reality seems to be we’re going to a state where you’ll have a choice of headless boxes (Mac mini and Mac Studio for the time being) to go with Apple-branded displays. In this scheme, one can upgrade the box without having to dispose of the display — which points to the problem I’d have if I wanted to update my own five-year old iMac, I’d need to jettison a perfectly good display in order to move to an Apple Silicon CPU. There is always the 24” iMac but, for the time being, it only sports an entry-level M1 chip.
A slightly orthogonal thought: because every word, every image of Apple Events is carefully vetted, one had to notice all six developers brought up to discuss the Mac Studio and Mac Ultra benefits inter work were women.
The Mac Studio and its M1 Ultra engine start shipping this coming Friday march 18th. That’s when we can expect the first reviews and reactions from the kommentariat. Right after the M1 Ultra introduction, a learned friend suggested the upcoming reactions from competitors might be an opportunity to update a September 2013 Monday Note titled 64 bits. It’s Nothing, You Don’t Need It. And We’ll Have It In 6 Months. In it, I joyously skewered naysayers who saw Apple’s 64-bit A7 processor powering the new iPhone 5s as a mere exercise in markitecture, as opposed to a portent of great things to come as the new chip “desktop-class” performance clearly signal a future beyond iPhones and iPads. In particular, we’ll have to watch Intel’s reaction, direct or telegraphed.