Apple Execs — How M1 Pro / Max Unleashed the Mac

Tom Boger (TB): Tom Boger. I run product marketing for the iPad and the Mac.

Tim Millet (TM): Tim Millet, and I manage the platform architecture team at Apple.

Rene Ritchie (RR): I figure at some point a few years ago, a decision was made that you were going to bring Apple Silicon to the Mac, but not just like the ultra low power Macs like the Air, I think the ones that everyone figured you would, but the ultra high performance Macs including these new Pros. What went through your heads at that moment?

TM: My first thoughts were, wow this is going to be a really interesting, exciting challenge for me and obviously for the team that I'm here representing. We knew that we had a lot of great fundamental building blocks in the Apple Silicon we were building for our IOS products, but the Mac is a very different beast, especially our Pro Macs which is what this recent launch was all about. And so, for that, we knew that the challenge wasn't to reinvent those fundamental building blocks, because those were in a great place. It was really to figure out how to scale them up, reconnect them, sort of take apart the IOS based solution and really take those fundamental building blocks put them back together in a way that was going to deliver the awesome performance that we were able to achieve with M1 Pro and M1 Macs.

TB: I was just tremendously excited. I knew what Apple Silicon could do. We obviously we're already demonstrating that with our, you know, most compact and affordable systems, and I could not wait to bring the ground breaking architecture of Apple Silicon to our higher end Pro products, because I knew it would be profound. I knew that it would completely change the game for our users, and just allow them to do things that they quite honestly couldn't do before, and I just couldn't wait. Tremendously excited.

RR: I've been using the new Mac Pro since just after you announced it. And, it's been, personally, the biggest breakthrough in traditional portable computing since that famous second generation MacBook Air, but at the opposite end of the spectrum. And, I've been doing workloads that used to routinely, consistently just thrash my old I9 MacBook Pro, turn it into this laboring blast furnace with fans as loud as a heli-carrier, and the new M1 Macs, it just stays utterly, almost eerily silent. And, I'm throwing everything I can at it, 4K video, a bunch of gnarly Canon Codex, processing audio, thumbnails, running Electron apps. And, I did that for 10 hours straight yesterday on battery, and I had 40% left in the tank when I had to quit. And, I know Tom, when you said that Apple Silicon would let you make Macs that just weren't able to do before, that weren't possible before, I think now that you were underselling how transformative it would be.

TB: Yeah. No. It's pretty incredible when you use these systems. And, what I find most interesting is that we've been taught as users for many, many years that, you know, you can't expect to have great performance without the fans going on. You can't expect to have great performance when you're doing Pro workloads when you're on battery. And now, in an instant, you can. It's so fun to watch all the videos that people are posting where they're trying this and they're trying that, and their mind is just blown about what they can do. They jump out of their chair, and literally can't believe what they're seeing. That is so fun for us here at Apple. You know, that kind of reaction, and that kind of just experience that we're giving people is why we come to work everyday and work as hard as we do. And, we love it, and we love the Mac, and so we're experiencing the same thing. We're doing things on our Macs that we can't believe we can do, and it's just really, really exciting time.

RR: Intel and AMD have forever been trying to cram these big thirsty server and PC tower parts into laptops, trying to scale them down to fit those much more constrained power and thermal envelopes, like trying to push a river through a very high pressure hose. But, you all started with the iPad and iPhone, and it seems very deliberately that you've let efficiency drive performance. Can you talk a little bit about your, you know, famous slow, low, wide approach, what it is maybe first, but also what it means when you now have the envelope the size of a new MacBook Pro to let it run in?

TM: This journey started for me back for the original iPhone. So, I've been working on every chip that's gone into every IOS product from the first iPhone, and you're right. That point, it was about how can we make this device more performant than anybody imagines it possibly could be? It was about delivering a level of performance that was surprising. And, because you had a touch based UI in those systems, you had no error bar there, really. If there's a laggy experience for the touch based thing on an iPad or an iPhone, the user notices it. And, the magic disappears, and so this is where the foundation of the obsession with delivering performance within the constraints of these ridiculously small devices came from. And, over the years, we had a lot of energy. We grew our team. We established some foundational architectural principles. And, energy efficiency was always that piece. We don't ever talk about performance without also talking about the energy that goes into delivering that performance. So, when we did get to the point where the fundamental building blocks, our CPUs, our GPUs, our neural engines were reaching the performance levels, they were competitive. This is when this discussion around the Mac became super interesting. And, you know, as we do with every product, this did not come about because we said, "Hey, let's build a chip idea and go pitch it to the Mac team, and see what they think." This was, "Hey, let's all get together and build and amazing Apple product." We built a chip of course, but we did it within the requirements, within the constraints of these amazing new platforms. And, they built it together. I kind of like to think that the new MacBook Pros, the 16 and 14 inches, they're kindly gently hugging these chips. Like, they're embracing them. They're built right in. They were not some part that we built and then had to wonder what the enclosure was going to be like. We built them in a very targeted way. And, like you started to say, efficiency delivers performance. Well, because the fundamental building blocks had that potential performance, we're finding ourselves delivering amazing performance inside these new devices. And, you know, you don't hear the fans go up because two things. It's a wonderful mechanical product design with, you know, amazing engineers doing that work. Plus, the chip, all CPUs maxed out, reaches a max power of around 30 watts. You saw that in the keynote slide. Well, 30 watts is not going to make the fans move. So, you can beat it up all you want with the CPUs. It's not going to do anything. But, that's not the only engine. We know we put a huge GPU. Our memory system is beyond belief in these systems. And so, why is it harder to come the other direction? Absolutely, and why? Because they were designing without those really fundamental constraints that we had to design to in the phone and the pad.

TB: These new Mac Pros, they're the ultimate expression of something that we've said at Apple for many, many years. And that is, "We build the whole widget," right? And, in this case, right down to the silicon. And so, to be designing something from the ground up where the architecture team is working on silicon specifically for these products and no others, this is unheard of in the industry. And, it's one of the reasons why these are just such game changing products.

RR: You have spoken before about your scalable architecture and how you know have an S7 based on an A13 E core in the watch. You have A15 in the iPhone and iPad mini, M1 and now M1 Pro and Macs in the Mac. And, that's like three generations of IP across four product lines and maybe five distinct SOC scales, but you're scaling this architecture from like four P courses and four or five graphics cores on an iPhone. Up to eight P cores and a ridiculous 32 graphics cores and 64 gigabytes of ram on the M1 Max. But, I also know that you and the team aren't just out there with like hot glue guns pressing all those new extra cores onto the die. What were the challenges in actually keeping all those cores fed, those pipelines full, but still scaling up Pro silicon for all the Pro workloads that, you know, Tom knew you had to deliver?

TM: It's great that you mention the Pro workloads, because that was sort of our fundamental motivation. When we were trying to figure out how to set the targets for this chip, you can imagine that it's a wide range. We wanted to figure out how with the tools we had to max out these new enclosures. So, it started with those Pro workloads. We worked closely with the Mac Pro workload team. We worked closely with the product designers to make sure we understood from, you know, from a strategy of IO. Let's make sure we have the right IO strategy around these chips. That's a big difference on these chips. You know, we don't have the ability to drive three plus one displays in any of the other chips. We had to go figure out that story. The scalability getting from where we were in M1 even to where we are in M1 Pro and M1 Macs was a fundamental re architecture of what we call the fabric, the interconnect through which all of these different cores connect together. And, the big challenge around unifying memory at this scale, because nobody's ever built this. There are unified memory examples all over the place. Every IOS system is a unified memory. Nobody has built unified memory at this scale. You talked about that 64 gigabytes. Well, along with that goes a 400 gigabyte per second memory system. So, why do we need such a crazy memory system on a computer like this? You need it, because that's the only thing that's going to keep that GPU fed. High end GPUs that operate in this class, they have memory systems this big. We knew we were going to need something like that if we were going to compete with the discrete GPUs that go in these Pro laptops. On the CPU side, the other thing that's critical, CPUs care about latency more than anything else. So, we had to figure out how to deliver that bandwidth to the GPU, low latencies to the CPU, but what falls out is incredible. And, you heard us talk about this in some of our previous briefings. You end up with a high capacity unified memory system, because a CPU needs that depth. It needs that 64 gigabytes. Pro users need the extra capacity. But, never before has a GPU been able to get that much access to a such high capacity memory system at that high bandwidth that it needs. And, never before has a CPU cluster gotten access to such a high bandwidth memory system before. I mean, it really is an interesting side effect of this, the way we built the machines. And, it was very intentional, because the GPU needs the bandwidth. The CPU needs the capacity. It also needs low latency. And so, the constraints are there. It's all over the place. And, putting it all together was a complete new redesign of the inner connect, the fabric and memory system from the M1 for example.

TB: And, you know that re architecture of the chip with the tremendous bandwidth and low latency of the unified memory architecture, it literally allows people to do things they couldn't do before. PC laptops, they top out at 16 gigabytes of memory. You heard Jules from Otoy talk about there's never been any hardware that allows a GPU access to 64 gigs. And so, when Truthy was talking about that topic, there was a scene behind her on the slide which is a model that our Pro workflow team created in Octane that has 137 million triangles. It requires nearly 35 gigabytes of video memory. And, on a 16 inch MacBook Pro with M1 Macs, you can interact with it fast and fluidly even in HDR. So, it's an HDR asset. And so, you can see HDR on that amazing screen. And, it is literally a workload and a workflow that was not possible before on a laptop. And, oh, by the way, you can do all that in battery. Which is just incredible.

RR: You both mentioned the Pro workflow team. And, for people who aren't familiar with them, you have essentially world famous award winning 3D artists and film makers and audio engineers, just the best of the best. People from Pixar, ILM, any place you can think of with high like super user people there. What does it mean for you and your teams to have sort of the pros of the pros hitting all of this stuff not after you release it, but like during the whole design period?

TM: Oh, it's fundamental to how we build things at Apple. As a person who's been building chips for more than 30 years now, doing that at Apple provides a level of focus that most folks who do what I do don't have. You know, I like to say that, "My team never has to guess." We never have to guess what software is going to run. They never have to guess what enclosure that we're going to be building into. We have certainty. It means that we can be really efficient in our focus. We can target things in particular and know that we don't have to target other things. And, when we make a decision, it's made years before it gets into the product's hands, which hopefully in the end, and I believe that's true folks when they get these products and they open them up, they really appreciate the consideration, the choices, that the intent was there. There's no surprises. Things work the way they're supposed to work. And, we don't leave anything on the table. We try to make sure that when we're building these chips we're using every transistor with intent, with purpose. We're not wasting transistors guessing whether something's going to be useful or not. And, because of the way we work with our partner teams across Apple, we don't have to guess. We know.

RR: It's your job every year to deliver the absolute best, formerly IOS, now IOS and Mac OS and app performance on the planet given the limits of time and the thermal envelope of the enclosure. And, schedules and industrial design can change a little bit, but physics like silicon is totally unforgiving. So, it seems to me, we're seeing some of the hard limits to just goosing frequency and ramping up voltage in the industry, but Apple seems to be increasingly focusing beyond just the big compute engines. Like you basically, this still blows my mind, you've basically taken a 2019 Mac Pro with Afterburner and Pro display XDR and you've shoved it a fairly mid to high end version of that Mac Pro into the new MacBook Pro. And, I know that's the definition of non-trivial, even beyond the big media engines, you're doing just a ton of work on performance and storage controllers, display engines, the fabric you mentioned. So, I guess what I'm asking you is when you talk about your transistor budgets these days, cores versus other features. How and what do you choose to spend it on? What sort of guides you to let you put all of that into this?

TM: I think there is absolutely a lot to be said about things like Afterburner. That's a great example about ProRes. ProRes acceleration, you know, why bother putting that down? You could run that on the CPU. And again, this goes back to we knew through our Pro workload team that this was going to be a super valuable thing to accelerate. And so, when we look at something like ProRes, why would we go accelerate it? What are the reasons you'd do that? And, it always comes down to two fundamental reasons. One of them is you have a constrained enclosure, and you know that through doing what we call a fixed function approach to something like a video codec like ProRes is going to allow you to do things that have significantly reduced energy. You save a lot by doing that. So, from an efficiency perspective it's always a big win. The second thing that falls out though is you get a big performance bump. You know, it was it necessary to do that? Was it necessary to achieve the performance? And, when we do the math on looking at our amazing GPU and our CPU, the answer is probably no. We probably didn't need to do the dedicated work, but when we looked at the energy equation, we said, "You know what? We don't want Pro users to have to worry about being plugged in when they're doing some of this work. We want to be able to do this work at an energy level that's more aligned." And so, when we do that trade off, we look at the investment in silicon. And, the third piece that sort of contributed to that amazing ability to kind of out perform the Mac Pro, the old Mac Pro with the Afterburner card goes back to your questions around SOC and the unified memory. We built those engines that went into the Afterburner card. We built the, we customized those engines, those ProRes engines. When we moved them and integrated them into the SOC, you saw dramatic improvements. And, that's because you moved them closer to the memory system. You got them closer to this shared 400 gigabyte per second memory system. And, that makes all the difference for this kind of a workload. So, it was a win, win, win. And, those are the easiest things for us to go and make our judgment call. And again, we sit down with our marketing team, with Tom's team. We sit down with the Pro workloads team, and obviously we sit down with, you know, my boss Johny Srouji who you all saw in the keynote. You know, when he sees a story like that, it's a pretty easy decision. We say, "Yeah. We're going to go do that."

TB: And, the other benefit for integrating things like the ProRes accelerators is it offloads that task from the CPU and GPU. So, they can do other things. We said it in the keynote. This is one of the huge advantages that we have in designing our own silicon specifically for our systems.

RR: For me, one of the biggest changes is yes, your ProRes rendering is like 10 times faster for me now. More than that, you know, it's giving me large chunks of my life back. But, it's every little thing is faster when I pull an effect it doesn't wait a few seconds every time. When I switch apps, it doesn't wait a few seconds. It's all those seconds off of all those minutes that's effectively given me a large part of my life back, and, I think for Pro users, that's an indescribable feeling.

TB: It's also unleashing you from being in a specific place, right, to be plugged in. You're getting that incredible performance and that incredible capabilities on the go on battery. And, I keep coming back to that, because if you're not familiar with Pro workloads and these kind of applications and ways in which our users use these systems, you don't realize that. You know, a lot of people who aren't working on Pro things, they don't realize that well, in this space, you really have to be plugged in. That's not true anymore, right? You know, we literally, quote unquote unleash people from being chained to being at a desk plugged in, because creativity strikes our users anywhere and everywhere. Could be on a train, at a cafe, in a park, a photographer who's in a tent because of shooting on location. Could be in the back of a taxi cab, and so, literally, I mean that happens. And so, you know, we want to make sure that people can have this amazing experience wherever they are.

RR: And now, that whole unleashed tag line makes all the sense in the world. I say this only semi in jest, but about the only complaint I've seen so far is that people no longer get a free space heater with every Mac. Your coffee doesn't stay warm. Your cat doesn't come over and cuddle. You have to buy a separate warm blanket if like me you live in Canada. That's a big change too, because it is, like I said, eerily quiet, eerily cool, and utterly responsive all the time. And, I can't imagine how many little things go into making that big of a difference. But, it's really impressive when you take it together.

TM: You know, I like to tell this story to folks on my team. You know, when I worked at other places in Silicon Valley building chips, the way things started is you'd build the chip. And then, you'd go figure out how to put it on a motherboard, and then you go figure out how to put an enclosure around it. And then, somebody figured out how to build fans big enough to keep the thing cool. And then, some software showed up. And, this is how systems conventionally have gotten built. Apple has completely upside down, and in my first few years, it took me a little while to figure it out. We start with a work of art imagined by our industrial design team followed by the best mechanical product designers in the world trying to figure out if such a work of art can be built with materials known to man. And then, the electronics team gets their budget. The software team chimes in about requirements, and the people, the lowly silicon people at the bottom are handed this sort of, "Hey, we need you to go off and figure out how to build a chip that allows this to be amazing." And, frankly, this is some of the best constraints that a chip developer could ever have. We are in a great position because of those constraints to build that experience that you talked about. That things that sort of life changing experience for these products, and that's in the end what it's about. It's about, not just silicon to kind of prove how amazing we are. It's silicon and service of delivering product, and delivering life changing experiences. And, it could not have been done at a place that didn't have that Apple culture. We couldn't have done this breakthrough innovation with these new MacBook Pros if we hadn't started with those fundamental constraints that led us to build silicon the way we do.

RR: I mentioned the MacBook Air, like that was the big revolution for me last time, but that just got so widely copied. That became the template for every ultra book. I don't know how you copy this. I don't know what anybody else does in the industry, because you're not a merchant silicon vendor. You don't have profit and loss per chip. You don't have to worry about maximizing shelf life to recoup the cost of engineering these chips. Your silicon team is probably the only one in the industry that is really just let loose to run generation after generation, year after year.

TM: Let loose, you know you've met my boss, Johny Srouji. He is a very disciplined guy. And, from my perspective, I think that's the other really important constraint that really drives this organization is that focus, that discipline, never waste a transistor. You know, even if we're not trying to make margin on that transistor, that's okay. Why are you putting a transistor down that you didn't need to?

RR Tom, I want to ask you a little bit about gaming, because it comes up a lot, and not just Shadow of the Tomb Raider benchmarks. Obviously, a ton of people at Apple are hardcore gamers. The iPhone and iPad are huge gaming platforms. You have Apple Arcade. There's Catalysts. There's Metal. But, for people who think that maybe Apple doesn't get gaming, or doesn't care about it, what's your current take on gaming on the Mac? Specifically, maybe these new MacBook Pros?

TB: We obviously definitely care about gaming from Apple Arcade and all of the great games that are on our various platforms. You know, these systems make a great gaming platform. And, we're excited about opportunities for developers to have great games on the Mac and bring great games to the Mac. Just like with Pro workloads from a gaming standpoint, they're, you know, systems that you just can't find anywhere else, and they're unlike any other notebook. And so, we think that this is going to be a great opportunity for developers to bring games to the Mac.

RR: So like from my point of view, just lobby your favorite studios. Just tell these studios what these things can do, and then demand that they bring all their games to them.

TM: Sounds good.

RR: Are there any little implementation details? Small things maybe that turned out better than you expected? Or, that you're just especially pleased or proud of? Like, personal favorite things when it comes to the M1 Pro and the M1 Macs?

TM: I gain great satisfaction out of our ability to predict many years ago, and as you can imagine, this has been a long journey. We don't, Apple doesn't just turn the whole Mac product line on a dime, and it doesn't do it very often. Frankly, I think going backward in time, it really wasn't a priority for us. Like we didn't say, "Oh, someday, we're going to take the Mac." We had a great product line. Intel was a great partner for the Mac transition as we went from PowerPC into intel. We built fantastic Macs. I think we're proud of every Mac we've ever built, but really what I think I'm most proud of is when we realized there was an opportunity to do something very interesting here, and we made our predictions about what we could deliver, the team and the collective team. Not just my team, but the broader Silicon Team at Apple was able to execute on that vision and deliver something that really was in the end, aligned with the predictions that we made to the company, to my boss, to the executive team at Apple so that they had that. They had faith in us and we were able to deliver on the vision that we were quoting

TB: As I mentioned, we love the Mac at Apple. We love the Mac. And, we want the Mac experience for every one of our users to be the best it possibly can. And, the reason is because we know that for our users in many cases, the Mac is their life. Their livelihood is on the Mac, right? And, that is, you know, you've heard that phrase before, you know, "You're not going to take this Mac out of my hands." And so, we know how important the Mac is to our customers. And, they love the Mac too. And, we want it to be the best it can possibly be.

RR: When you look at things like the port arrangements, you know, you went with Thunderbolt 4, which a lot of people love. There's HDMI 2.0. How do you figure out the bandwidth that you have for all of the connections and then choose, like we're going to go all the way with Thunderbolt, but people aren't really connecting big displays on HDMI. That's mostly a TV thing, so we're going to leave that bandwidth on the Thunderbolt ports. Is that all about the trade offs, or the workloads that you see in every day life?

TM: We start our chips, of course, many years in advance. And so, the nice thing about some of these IO choices are we can get unified on the kinds of bandwidths and set our specifications accordingly. And, he system team has a little more time to figure out the exact arrangement, what kind of port arrangements do they want to have? What kind of bandwidths? You know, as long as we are showing up with a chip capable of driving a reasonable set, a reasonable arrangement of ports across different systems, I think, you know, that's how it translates for us. I think the important choices that the system team makes about that are driven by all those great Pro inputs we get from the work flow teams, from discussions they have with customers, from Tom and all the work he does in outreach to try to make sure we're hitting the nail right on the head for the Pro customer.

RR: We now have these new MacBook Pros with everything from, you know, 6P cores and 14 GPUs on the 14 inch to 32 GPUs and a high power mode on the 16 inch, you know, talk about your love letter. For both of you, what is your go to?

TM: I am super excited that we were able to put M1 Macs into the 14 inch, because I love the portability of the 14 inch. I love that you can pick it up, carry it around, flip it open, do a bunch of work. And, the fact that we were able to put that massive computer in there, that is a point of pride to me. It really is the ultimate love letter to the pros that prefer that portability. You know, I couldn't be happier with how the 14 inches turned out.

TB: One of the great things about that system is and I'm not sure a lot of people have realized this. But, when we introduce these new MacBook Pros, we actually introduced a whole new product category for the Mac, and that's that 14 inch MacBook Pro. If you look back on our history of the MacBook Pros that we've been offering, there's been quite a distance between a 13 inch MacBook Pro, and, you know, years ago, a 15 inch MacBook Pro, and now more recently a 16 inch MacBook Pro. They've been very different animals, right? This is the first time ever where we have this amazing 16 inch MacBook Pro with these amazing capabilities, and now we have the 14 inch MacBook Pro which is really just all of that with a smaller screen.

RR: I was really tempted by the 14 inch, just because of how mobile it is, and when the world stops ending, and I can go back to traveling I think that is so tempting, but at the end of the day, I live in Final Cut Pro, and having the ability to have basically a Mac Pro with Pro Display XDR all that screen real estate, all those pixels in front of me to work with, to make it my main editing machine, I just couldn't turn it away. And, when you combine the battery life that you get with the 16 inch, it really does feel like a Mac Pro book as much as a MacBook Pro. So, I'm all in on the 16 inch right now.

TB: Well, we appreciate that too. It's that dedication, that passion that we have for the Mac. Hopefully, our users are experiencing in these products. And, you know, these are awesome, awesome systems, and, you know, I personally having been around quite a while working on the Mac, have never been more excited about the future of the Mac.

TM: And, I have to say, you're words are gratifying. And, I love to be able to speak to folks like you who can tell that story so well. And, you know, you are a user. You experience the goodness of the machine, and when you talk about it, I love to bring that back to my team and the teams we work with, and say, "Look at the difference you make," because you're right. There's a lot of our lives that are dedicated to doing this engineering. We love it. It's hard work, and it really makes all the difference when we get that product out there and the customers start going, "Wow, how do you do it?" So, thank you for the kind words, appreciate it.