Apple’s M series The arrival of chips in late 2021 was extremely well telegraphed. Apple had been making its own silicon since the A4 appeared on the iPhone 4 a decade ago. The appearance of Apple’s in-house efforts in the Mac was really just a question of when, not if.
It landed with a powerful bang when the M1 arrived. The chips were regarded as a significant step forward in portable computing, not only because they were noticeably faster but also because of their astonishingly improved performance per watt. This allowed for full-speed processing on battery power while allowing for increased usage times and a longer life.
Apple has just launched the next iteration of the M line with this year’s M2 MacBook Pro and Mac mini models — officially denoting this as an ongoing series rather than a one-off leap. With confirmed 20% improvements in CPU and 30% in GPU performance in under 2 years and a really aggressive entry price point, the M2 adds to Apple’s lead in portable chipsets.
I was able to spend a bit of time talking to Apple’s vice president of Platform Architecture and Hardware Technologies Tim Millet, as well as VP of Worldwide Product Marketing Bob Borchers about the impact of the M chips so far, how they see the line developing over time and a bit about gaming too.
Resetting the baseline
“A lot of it comes down to the people and the talent on Tim’s team,” says Borchers, “but I think a lot of it also comes down to the way that we’ve approached designing Apple silicon from the very beginning.”
Millet, who has been working on chips for over 30 years, has been with Apple for almost 17. Millet says Apple saw M1 as an opportunity to “really hit” it.
“The opportunity we had with M1 the way I looked at it, it was about resetting the baseline.”
When their desktop computing and laptop computing pipeline was essentially controlled by the third-party merchants and silicon vendors, it didn’t really allow for Apple to push the bar closer to the limits of technology.
This need to own and “reset” the baseline of portable performance in computing coalesced around the time that Apple started working on the iPad Pro. They had been building chips inside of these super-thin enclosures and knew that, with ready power and much larger casings, they could make a significant impact on portable computing.
“Once we started getting to the iPad Pro space, we realized that ‘you know what, there is something there.’ We never, in building the chips for iOS devices, left anything on the table. But we realized that these chips inside these other enclosures could actually make a meaningful difference from a performance perspective. And so with M1 we were super excited about the opportunity to have that big impact — shifting all of it back up to redefining what it meant to have a laptop in many different ways.”
“The M1 Opportunity we had, the way I saw it, was about resetting our baseline.” Tim Millet is Apple VP
The work that Apple did with the M1 wasn’t focused on pure peak performance, says Millet. From the beginning, there was this idea that they’d be able to reset user expectations around what kind of performance you should be able to get out of a portable computer, and for how long. The focus on performance per watt paid off (as noted in my early review of Apple’s first M1 chip), in that people could run major compute tasks on laptops untethered from power for Hours. No compromises. That, says Millet, wasn’t a byproduct — it was the intent from the beginning.
“We wanted to have the ability to build a scale of solutions that deliver the absolute maximum performance for machines that had no fan; for machines that had active cooling systems like our pro class machines. We wanted to…move performance per watt to the point where we delivered real usable performance in these in a wide range of machines.”
Tim Millet, Apple vice president Platform Architecture and Hardware Technologies
Millet claims that Apple was happy with the M1 product it was able to ship and that it met its goals. After the machines were shipped, my experiences and those of power-hungry customers echoed that sentiment. For decades, Apple had been running up against third-party stewardship over chipset speeds, power requirements and features — with the result of increasingly less mobile computers that ran hot, loud and short. The whole portable industry had been constricting that flexibility for so long that most users (aside from those of us who spend our lives closely examining these boundaries) probably didn’t realize how hard it was for them to breathe.
The M1 pushed a reset button on these restrictions, which put portable back in the power computing lexicon. Millet states that Apple didn’t want to extract a few percentage points from each generation for ever.
“The M2 family was really now about maintaining that leadership position by pushing, again, to the limits of technology. We don’t leave things on the table,” says Millet. “We don’t take a 20% bump and figure out how to spread it over three years…figure out how to eke out incremental gains. It’s not that we do it all in one year, but rather, it’s a very hard hit. That’s not what happens in the rest of the industry or historically.”
Borchers comments that Apple builds products and not parts. This allows it to close the loop between deliverables and needs. He notes that the pairing of the technology and product is not an independent choice at Apple; it’s silicon, software and hardware coming together, starting at the point of inception. It’s not “Can we do this?” Waiting to see if the vendor can deliver the right capability.
“As somebody who’s been building silicon for 30-plus years, the luxury of knowing what the target is, and working side by side with the product designers, the hardware system team, the software people to understand exactly what you’re aiming at, makes all the difference in our ability to really target and make sure we’re adding things that matter, not adding anything that doesn’t,” Millet agrees.
Forcing fans to perform and forgetting about them
The Mac’s modern history has seen Apple lose control of its machine capabilities. In 2006, Apple switched from PowerPC to x86. It had to include an external factor, such as Intel, to achieve the goals it set out to achieve with its new machine. This allowed it to deliver on their timelines and complexity. There was a force function. The partner says to you “This is the best we can deliver you on this scale in this timeframe. This goes all the way through product design, development and production.
Apple retakes control over its silicon pipeline Also I have reset the complexity and size of that feedback loop. This prompts me ask them if that relationship has changed at Apple after M1.
“If you go back to the phone, we had that tight interaction really, from the very beginning — and I think that’s true for all of the iOS products. We had that tight feedback loop,” says Millet.
Both Millet and Borchers are diplomatic about the Intel partnership (which is till present, for now, in Apple’s Mac Pro machines).
“Intel was a great partner through the years where we shipped the Intel machines. They were responsive and inspired by the directions that Apple gave them. This interaction I believe benefited our products. Of course, our competitors’ products benefited from that interaction as well sometimes,” notes Millet.
The relationship between Apple and what Apple does is undoubtedly strong. wants What to ship? can The M-series chips have been added to the breadth and depth of the ship’s lineup. Apple’s chip team working closely with internal teams has been a natural part of Apple’s device pipeline since it began work on the iPhone 4. This system now encompasses the Mac branch.
“I think it felt very natural for us to sit down side by side with our industrial design partners and our system team partners inside Apple because they’re familiar faces to us,” Millet says. “These are people that we’ve been working with for iPad and iPhone. It was very, very natural. It’s a very Apple way of working where we are all sitting at the table together, imagining possibilities and them challenging us and us going back and doing the math to kind of figure out that ‘yeah, I think we can do that without a fan.’”
Borchers sparks on my use of the term “feedback loop” and notes that it’s less that the loop is smaller and instead that it has been eliminated.
“[That term] implies some sense of latency or delay in the cycle,” he says. “And I think that it’s an appropriate way of thinking when you’ve got multiple parties involved. I think that the big difference here is that we move from having a feedback loop to co-creation, where there isn’t a feedback loop…You [just] Sit down at a desk and push one another. What if the fan was removed? This doesn’t mean that the fan has to be removed. It also allows for more creativity and efficiency. This is where I think you have hit upon something very interesting. It’s the difference between feedback loops and feedback. Different kind of process.”
M2 Max has 67 billion transistors and 400GB/s of unified memory bandwidth. It also features up to 96GB fast, low-latency, unified memory. (Image: Apple)
Gaming on the Mac
One arena still holds fascination for any of us who have found a home on the Mac for nearly every part of our digital life — save one: gaming. Due to the inclusion in the M-series Macs of highly-improved GPUs throughout the line, they are certainly more capable gaming than any other Mac. However, even though there are many big titles appearing on Mac in spurts there is still a substantial section of “here come the dragons” where Apple would like to map out the multibillion-dollar gaming market.
Borchers believes that Apple feels like its silicon gaming story is becoming more solid with each release.
“With Capcom bringing Resident Evil across, and other titles starting to come along, I think the AAA community is starting to wake up and understand the opportunity,” he says. “Because what we have now, with our portfolio of M-series Macs, is a set of incredibly performant machines and a growing audience of people who have these incredibly performant systems that can all be addressed with a single code base that is developing over time.
“And we’re adding new APIs in and expanding Metal in Metal 3, etc. And then if you think about the ability to extend that down into iPad, and iPhone as well, I think there’s tremendous opportunity.”
He acknowledges that Apple needs to do work to bring game developers along the road to adoption, but he says the company is happy that they’ve shipped the core ingredients in very performant systems. He said that the team will continue to examine a wide range of chips configurations and components from a gaming perspective. Millet believes that anyone who plays on Macs should be encouraged by the team’s focus here. Time will tell.
Millet says that Apple’s work on cracking the gaming market started with the early days of the Apple silicon transition.
“The story starts many years ago, when we were imagining this transition. Gamers are serious people. And we don’t believe we can fool anyone by pretending that we will make Mac a great gaming platform in a matter of hours. We’re going to take a long view on this.”
He notes that Apple offers common building blocks that are shared, scaled appropriately, between the Mac, iPhone and iPad where Apple has historical strength. But he also points out that the purpose-built GPUs in the iOS devices weren’t intended to be general use.
“We weren’t going to design GPUs for that space that were unnecessarily complicated, that had features that were not relevant to iOS,” he notes. “But as we looked at the Mac, we realized that this is a different beast. There will be different expectations over time — let’s make sure we have our toolbox complete.
“And so we did very directed work to make sure that the GPU toolbox was there — working super closely with our Metal partners. We worked hand in hand to make sure that they were going to have all the tools that they needed to accelerate the important APIs that we’re going to deliver to [companies like] Capcom, for example. So that when Capcom approached us, it wasn’t going to be this awkward port for them. It was going to be a very natural ‘Ah, you do support these modern APIs that gamers are needing. This is interesting.’”
That, in turn, makes it simpler to approach game developers with a strong set of tools that feel familiar enough and compatible enough with their existing workflows to make porting games or building for the Mac viable.
“My team spends a lot of time thinking about how to make sure that we’re staying on that API curve to make sure that we’re giving Metal what it needs to be a modern gaming API. We know this will take some time. But we’re not at all confused about the opportunity; we see it. And we’re going to make sure we show up.”
He also acknowledges that it will take time to build an installed base of strong GPUs in order for it to be enticing to the AAA space.
“The other thing we wanted to do, and I think we have hopefully done, is to seed the Mac, the full Mac lineup, with very capable GPUs, whether it be the MacBook Air, obviously, all the way up to the beast, Ultra chips that we can put in our Mac Studio.”
“Because until you do that, until you have a population distributed, developers are going to be wary about making a big investment and kind of focus on Mac,” Millet acknowledges.
So Apple will continue to seed the Mac population as people upgrade from Intel to M1 or M2, and it will, hopefully, become more and more obvious to developers that the Mac population at large has a machine that is capable of running major titles at a frame rate that is acceptable to gamers.
Millet also is unconvinced that the game dev universe has adapted to the unique architecture of the M-series chips quite yet, especially the unified memory pool.
“Game developers have never seen 96 gigabytes of graphics memory available to them now, on the M2 Max. I think they’re trying to get their heads around it, because the possibilities are unusual. They’re used to working in much smaller footprints of video memory. So I think that’s another place where we’re going to have an interesting opportunity to inspire developers to go beyond what they’ve been able to do before.”
The custom technologies of M2 Pro and M2 Max include Apple’s next-generation, 16-core Neural Engine and new powerful, efficient media engines with hardware-accelerated H.264, HEVC, and ProRes video encode and decode. (Image: Apple)
Historically, the most vibrant chatter about the Mac is about the next Mac. No matter what gains or features Apple delivers with a particular system, the question readers and friends are always hitting me with is, “When is the next one coming, and is it worth waiting for?” Lest you think it’s just happening out here in user land, Millet says that he gets it, too.
“Friends and family reach out all the time and they say, ‘Hey, I’m thinking about getting a new Mac, wink, wink. Is now a good time?’ And what’s beautiful about this story is that I really, with full sincerity, believe now is always a good time…Nobody should be shy about it.”
“I don’t think we’re going to fool anybody by saying that overnight we’re going to make Mac a great gaming platform. We are going to have a long-term view of this. Tim Millet is Apple VP
The Mac with M2 onboard has received so many updates that it is almost impossible to not be impressed by the enthusiasm. The broad consensus, in fact, is that the M2 Mac mini is probably the best-value computer that Apple has ever shipped — and may actually be the best value in desktop computing period. It costs $599 for the base model. Lower price than the previous M1 model and is incredibly capable — clocking in at around 20% faster. Even with that nice positioning, however, the previous year’s model remains more than viable for all of the use cases in that band.
Apple has found a sweet spot where customers can “buy whenever” knowing that M-series chips will be available. It’s good!. Millet claims that Apple is aware that only a few percent of power-hungry users want to have the best possible capability. They are smart enough to know the approximate release date of the “new Macs” so they will wait.
“That 20% is going to make a big difference to some folks. Absolutely. They will be patient. [when] they can see it because they have workloads that require it,” Millet says, while noting that they’re confident enough in the satisfaction levels of buyers of M1 that they’re not going to hold off.
“If you bought a MacBook Pro last year with M1, you’re gonna be fine. [Even] If you bought it in December you won’t be screaming at me that you hate the machine. [and] why didn’t you tell me to wait?”
Apple wanted to maintain a consistent pace of shipping M2. This is one reason for shipping M2. It was important, Millet says, to make sure people didn’t see the M1 as a “one and done.”
As far as the “when Macs” question goes, Millet and Borchers are both in the “when possible, ship” camp. Coming out of a period pre-M1, when many in the Mac ecosystem felt that it was being underinvested in, it’s clear that Apple wants to send a message that this is not the case and they never want that to become a meme again.
“As a silicon person, I know that technology moves fast and I don’t want to wait around. I certainly want to push hard, as you can imagine,” says Millet. “We want to get the technology into the hands of our system team as soon as possible, in the hands of our customer as soon as possible. We don’t want to leave them wondering…do they not care about us? An all-new phone was shipped last year. Why didn’t the Mac get the love?”
“We want to reset to the technology curve and then we want to live on it. We don’t want the Mac to stray too far away from it.”
Borchers claims that Apple’s opportunity lies in the fact the vast majority are using Intel machines. This makes it less of an “it’s been a year, we have to ship something” situation.
“We’re just trying to make it more and more of an easy decision to move…to an even more amazing system,” he says.
Positioned year-over-year, the gains are very solid. If Apple is marketing to and eyeballing millions of Intel Mac users, the math (for them) becomes much easier. As proud as the team is of the M1 to M2 speed jump, Borchers says, the real messaging is around the leap from just 2 years ago — especially for the Mac mini, where the leaps are in the 10x and up multiples of performance. All in a price that’s $100 cheaper than the M1 Mac mini and $200 cheaper for students.
“We’re product people at the end of the day, and we want to put our systems in as many hands as possible,” says Borchers. “We feel like the Mac mini form factor is such a great way to unleash creativity and, frankly, goodness in the world that we wanted to be able to put it in as many people’s hands as possible.
“We don’t think about it.” [Mac pricing] in a traditional kind of cookie-cutter way where it’s like, ‘Okay, it’s 2023, we’re going to $799 and we’re going to be predictable.’ It’s more of what do we have [in the pipeline], and what can we do that will surprise and delight our customers?”