NEW M1 MacBook Pro / Max — Changes EVERYTHING!

Apple has just announced the new M1 Pro and M1 Max 14-inch and 16-inch MacBook Pro with mini-LED display — and notches! — HDMI, SD-card, MagSafe, fast charging, 8 performance cores, 2 efficiency cores, up to 64 GB of RAM, up to 32 GPU cores, and 120Hz ProMotion! And these are my live reactions to the event!

Oh. Yeah!

• This is M1 Pro.

• Holy wow. I'm Rene Richie. I'm reacting live. M1 Pro, they are going with the M1 Pro branding. So not M1X, like they used to do A12X, they're going M1 Pro.

• And it's a game changer. Now let's hand it over to Johny for a deep dive into this remarkable new chip.

• And now we're going to Johny Srouji. Johny Srouji joined Apple to start pushing out the first generation of Apple Silicon, the A4. He's been leading Apple Silicon division. He's now Senior Vice President of Hardware Technologies, which is basically all the internal stuff, the silicon, the modems, everything that goes inside the devices. And he's been spearheading Apple Silicon this whole time.

• [Johny] Now building up a laptop has meant using a power hungry CPU and this good GPU but the two-chip architecture requires much more power and cooling. It also means the CPU and GPU have separate pools of memory. So they have to copy data back and forth over a slower interface.

• So he's talking now about how traditional computers work. I've described them as a charcuterie board as a buffet where every part is laid out separately, and you have to communicate in between, and you got to waste time reaching for the different parts where an SOC is more like a sandwich.

• [Johny] And we did this by scaling up M1's groundbreaking architecture to create a far more powerful chip with M1 Pro.

• And now, wow, this is way bigger than M1 This is just a wide body.

• While doubling the width of the memory interface and using faster DR.

• 200 gigabytes of memory bandwidth, three times, as much as M1. That's amazing.

• [Johny] And its custom package supports up to 32 gigabytes of unified memory.

• Five nanometer process, probably five nanometers P at this point, maybe.

• [Johny] Starting with the 10 core CPU with eight high-performance cores and two high efficiency cores.

• No word yet. If these are A14 generation or A15 generation. The A15 generation are slightly faster, slightly better efficiency. And the efficiency cores are way faster. So we'll see.

• [Johny] 16 GPU cores, eight more than anyone.

• And we'll see if there's an even bigger an even bigger Pro Max version

• M1 Pro has our industry leading media engine that accelerates video processing while using very little power.

• Apple's really all in on this post pro post big cores.

• [Johny] Adding forwards acceleration enables M1 4 to play multiple streams of 4k and 8k ProRes video, while using just a fraction of the power.

• Oh yeah. And here's the ProRes. We saw this with the A15 ProRes acceleration was huge. It basically turned the iPhone into a mini afterburner box, and this isn't going to be a mini- a maxi after burner box. So you'll be able to do things on M1 Pro that you needed an afterburner card on the Mac Pro just two years ago to do.

• Finally M1 Pro includes other advanced technologies for a complete Pro system like a powerful engine to drive multiple displays and additional Thunderbolt controllers to deliver even more IO bandwidth and so much more.

• [Rene] So yeah, what else we have here? Secure enclave. Thunderball 4! Great! Oh, I was too slow. I'm gonna have to zoom enhance, zoom enhance. Does that pruner that diagram two new chips here go. M1 Pro and Johny gets to announce it that is so good. I've been waiting for this. M1 Max there we go.

• This incredible chip builds on M1 Pro and takes its amazing capabilities to new heights.

• And I like that they stuck with M1 because it does define the generation, these aren't next generation chips that go assumedly- presumably get an M2 with the next MacBook Air, But this is the same generation Just more. It's extended into what would be an X chip.

• M1 Max starts with a much higher bandwidth on chip fabric and doubles the memory interface once again.

• 400 Gigabits per second, twice M1 Pro, six times M1.

• To 64 gigabytes of unified memory. And its VI has a staggering 57 billion transistors. That's 1.7 times M1 Pro and 3.5 times M1.

• So basically M1 Pro Max it's like two M1's. It's like four M1's in terms of surface area. Not quite. Maybe 3.

• [Johny] M1 Max as the same powerful 10 core CPU complex of M1 Pro and doubles the GPU to a massive 32 cores.

• So it really is just part of Apple's scalable architecture strategy.

• [Johny] And it has an even more capable media engine with up to two times faster video encoding, and two ProRes accelerators.

• To ProRes and co- decode engines? Give it to me now. I want it now. And here's the performance per watt. The efficiency.

• [Johny] And when you add an M1 Pro and an M1 Max, they deliver dramatically more performance.

• Oh, Apple just redefined the curve. And now, some people are gonna get upset that these are bayzos graphs without anything, but 4-core laptop, oh, it's too small for me to read, but it's going to be whatever Tiger Lake, or I forget what we're on now I stopped paying as much attention to Intel.

• [Johny] At their peak, they delivered 1.7 times the performance of the PC chip in the same power envelope.

• So we had around 15 watts for the M1. This is around 30 watts.

• And at the peak performance of the PC chip, they deliver the same performance at 70% less power.

• Yeah. This is where I watched x86 die .

• [Johny] So, when we add an M1 4, it's in a whole different class. At its peak, it's more than seven times faster.

• There we go. And M1 Max, I want the graph for M1 Max. Hit me, hit me with that graph for M1 Max.

• [Johny] M1 Pro delivers more performance while using 70% less power.

• Usually, yeah, you'll have an Nvidia or an AMD discreet. Even Apple's previous generation of MacBook Pros had discreet, but they use a ton of power.

• [Johny] It delivers competent performance while using 40% less power.

• And that gives you a trade-off. Either you've got to feed it a ton of power, or throttle it down. And some of them, when you pull out the plug, they throttle all the way down. So it's never been a really good balance.

• M1 Max delivers similar performance while using a hundred watts less power.

• Like we're going to have to wait and see, we're going to have to review them. We're going to have to test them out. We're gonna have to try them with a diverse array of Pro workloads to see how, not just how well these GPU's actually perform, but how compatible they are. You know, a lot of people still want their kudo cores. A lot of people now are used to AMD GPU's. Apple wants everyone to write to metal. There's huge advantages to writing to metal.

• [Johny] If you look at the compact pro laptop on battery its graphics performance drop significantly. M1 Max is up to two and a half times faster. And if you look at the high end laptop when it's on battery, the drop off in graphics performance is even more extreme. M1 Max is over three times faster.

• The big battle between Apple and Nvidia has been that, you know, Nvidia wanted all the boxes to be commodity PCs. And they wanted to be the important part where Apple wanted, you know, all the graphics to be commodity parts and the important part to be the Mac, the metal. And they've been just bouncing, locking heads over this for a while now. And this is Apple's big play to sort of have their worldview imposed at least upon Mac users. We'll see how it goes.

• It's by far the most capable chip we've ever built.

• [Rene] Craig Federighi of course is Apple Senior Vice President of Software Engineering. He's the traditional superhero of Worldwide Developers Conference, WWDC, every year.

• And my personal favorite? Instant wake from sleep.

• So this is interesting: Apple Silicon team- They typically look at current and upcoming workloads. They see the type of thing that users are doing, the type of apps that are coming on the market, the apps that are expected to come on market and they try to predict the performance that you'll need the performance and efficiency that you'll need for those workloads. So there've been building these things.

• To help optimize how Mac OS assigns multi-threaded tasks to the CPU cores. Advanced power management features intelligently allocate tasks between the performance and efficiency cores.

• It takes about three years to fully build out one of these chip sets. So they've been looking at where they want it to be for these last few years.

• [Craig] And Mac OS is designed to take advantage of the unified memory architecture in these new chips. So Pro apps can manipulate huge images or video streams.

• And here we're talking about unified memory, which is interesting because traditionally you've had the Ram for the CPU and the V Ram for the graphics chip. And you don't have that with unified memory. It's shared memory, which does have some trade-offs of course, but it means that with these M1 Pro and M1 Max chips, that 16 or 32 core GPU is going to be fed by 16 or 32 gigabytes, or 64 gigabytes of Ram.

• We also have Rosetta 2. So Intel based apps, even pro apps with plugins simply run without missing a beat.

• They've got accelerators in there that basically reorder things so that they run as fast as possible. And if they've already been built for metal, they will sometimes run faster on Apple's architecture Than on the Intel x86 architecture, which is just mind-boggling.

• [Craig] With Logic Pro for the first time, musicians will be able to create massive spatial audio mixes on a notebook. In Final Cut Pro, video analysis for the new object tracking feature is up to five times faster.

• Because you know, all that silicon is great, but you got to have the bits to go with the atoms.

• ProRes Video Transcode is a remarkable 10 times faster.

• That's basically what I've been looking for is these boxes to become these little afterburner boxes or for whatever we're doing video or audio or code or, or whatever your 3D, whatever your pro workload is for these to become just massive accelerators, dedicated accelerators, not big cores anymore. The big core era era is probably over, because no matter how you know how fast they get there, they're going to hit eventually up against the laws of physics The size of the thermal envelope of a device that we, you know, that they just can't get endlessly big, but also the power that's available to them. And you don't want to spike power and brown out devices. So moving off core, doing things like H.264, HEVC, ProRes, encode decode blocks, spatial audio, audio unit processors, all of these things that aren't the GPU, the CPU, by pushing things onto the neural engine course and onto AMX, which are matrix multiplication blocks on the CPU. That just means that you can accelerate all these tasks without having to invoke the bigger, hotter, you know, greedier, thirstier cores all the time. And Apples are really efficient. Their big cores are really efficient comparatively, but you still don't even have to do them. That's also one of the reasons benchmarking is traditionally been- well, not traditionally, It's traditionally been pretty easy. I mean, people just download an app, press a button, and they call that a benchmark, but they've become really tricky over the last few years, because unless you're an expert, unless you're like on a Nan tech level 9,000 IQ, you often don't know if you're hitting an accelerator block, Like, are you testing two different iPads? Are you testing two different iPads or are you testing the same accelerator block in an A12X and an A12Z? Because those things didn't change. Is it hitting the GPU? Is it hitting the CPU? Is it hitting the neural engine core? All of these things are, if you look at any of Anandtech's reviews, these are non-trivial things and they're not well understood. So measuring them has been hit and miss at best and understanding how unified memory works. For example, some people just think that eight gigabytes of unified memory is as good as six gigabytes of traditional 16 gigabytes official memory. And it's not, it's just highly optimized, like because of, because of unified memory, because of memory compression, because of super fast swap, you get everything that you can out of that memory out of that eight gigabytes of memory, but it doesn't really help you save you from memory pressure when you're running two or three pro apps at the same time there you actually need the Ram. So Apple giving us, you know, 16, 32, 64 gigabyte options for that memory is going to be really important, especially for pros who want to have all the layers, all the objects, everything in memory it's, it's going to make a huge difference.

• M1 Pro and M1 Max represent a huge leap forward in Silicon for pro systems.

• Gonna have to wait for the complete review to see how they really live up to expectations.