Apple Announces the M1 Processor Powering Next-Gen Macs

[btarunr]

Apple today announced M1, the most powerful chip it has ever created and the first chip designed specifically for the Mac. M1 is optimized for Mac systems in which small size and power efficiency are critically important. As a system on a chip (SoC), M1 combines numerous powerful technologies into a single chip, and features a unified memory architecture for dramatically improved performance and efficiency. M1 is the first personal computer chip built using cutting-edge 5-nanometer process technology and is packed with an astounding 16 billion transistors, the most Apple has ever put into a chip.

It features the world's fastest CPU core in low-power silicon, the world's best CPU performance per watt, the world's fastest integrated graphics in a personal computer, and breakthrough machine learning performance with the Apple Neural Engine. As a result, M1 delivers up to 3.5x faster CPU performance, up to 6x faster GPU performance, and up to 15x faster machine learning, all while enabling battery life up to 2x longer than previous-generation Macs. With its profound increase in performance and efficiency, M1 delivers the biggest leap ever for the Mac.



There has never been a chip like M1, our breakthrough SoC for the Mac. It builds on more than a decade of designing industry-leading chips for iPhone, iPad, and Apple Watch, and ushers in a whole new era for the Mac," said Johny Srouji, Apple's senior vice president of Hardware Technologies. "When it comes to low-power silicon, M1 has the world's fastest CPU core, the world's fastest integrated graphics in a personal computer, and the amazing machine learning performance of the Apple Neural Engine. With its unique combination of remarkable performance, powerful features, and incredible efficiency, M1 is by far the best chip we've ever created."

First System on a Chip for the Mac
Macs and PCs have traditionally used multiple chips for the CPU, I/O, security, and more. Now with M1, these technologies are combined into a single SoC, delivering a whole new level of integration for greater performance and power efficiency. M1 also features a unified memory architecture that brings together high-bandwidth, low-latency memory into a single pool within a custom package. This allows all of the technologies in the SoC to access the same data without copying it between multiple pools of memory, further improving performance and efficiency.

The World's Best CPU Performance per Watt
M1 features an 8-core CPU consisting of four high-performance cores and four high-efficiency cores. Each of the high-performance cores provides industry-leading performance for single-threaded tasks, while running as efficiently as possible. They are the world's fastest CPU cores in low-power silicon, allowing photographers to edit high-resolution photos with lightning speed and developers to build apps nearly 3x faster than before. And all four can be used together for a huge boost in multithreaded performance.

The four high-efficiency cores deliver outstanding performance at a tenth of the power. By themselves, these four cores deliver similar performance as the current-generation, dual-core MacBook Air at much lower power. They are the most efficient way to run lightweight, everyday tasks like checking email or browsing the web, and preserve battery life like never before. And all eight cores can work together to provide incredible compute power for the most demanding tasks and deliver the world's best CPU performance per watt.

The World's Fastest Integrated Graphics
M1 includes Apple's most advanced GPU. It benefits from years of analysis of Mac applications, including everyday apps and challenging pro workloads. With industry-leading performance and incredible efficiency, the GPU in M1 is in a class by itself. Featuring up to eight powerful cores capable of running nearly 25,000 threads simultaneously, the GPU can handle extremely demanding tasks with ease, from smooth playback of multiple 4K video streams to rendering complex 3D scenes. With 2.6 teraflops of throughput, M1 has the world's fastest integrated graphics in a personal computer.

Blazing-Fast, On-Device Machine Learning
The M1 chip brings the Apple Neural Engine to the Mac, greatly accelerating machine learning (ML) tasks. Featuring Apple's most advanced 16-core architecture capable of 11 trillion operations per second, the Neural Engine in M1 enables up to 15x faster machine learning performance. In fact, the entire M1 chip is designed to excel at machine learning, with ML accelerators in the CPU and a powerful GPU, so tasks like video analysis, voice recognition, and image processing will have a level of performance never seen before on the Mac.

More Innovative Technologies Packed into M1
The M1 chip is packed with a number of powerful custom technologies, including:

• Apple's latest image signal processor (ISP) for higher quality video with better noise reduction, greater dynamic range, and improved auto white balance.
• The latest Secure Enclave for best-in-class security.
• A high-performance storage controller with AES encryption hardware for faster and more secure SSD performance.
• Low-power, highly efficient media encode and decode engines for great performance and extended battery life.
• An Apple-designed Thunderbolt controller with support for USB 4, transfer speeds up to 40 Gbps, and compatibility with more peripherals than ever.

macOS Big Sur Optimized for M1
macOS Big Sur is engineered, down to its core, to take full advantage of all the capability and power of M1, delivering a massive boost in performance, astonishing battery life, and even stronger security protections. With M1, things users do every day feel noticeably faster and smoother. Just like iPhone and iPad, the Mac now instantly wakes from sleep. Browsing with Safari—which is already the world's fastest browser—is now up to 1.5x speedier at running JavaScript and nearly 2x more responsive.2

With Big Sur and M1, Mac users can run a greater range of apps than ever before. All of Apple's Mac software is now Universal and runs natively on M1 systems. Existing Mac apps that have not been updated to Universal will run seamlessly with Apple's Rosetta 2 technology. And iPhone and iPad apps can now run directly on the Mac. Additionally, the foundations of Big Sur are optimized to unlock the power of M1, including developer technologies from Metal for graphics to Core ML for machine learning.

Start of a Two-Year Transition for the Mac
M1 powers the new MacBook Air, 13-inch MacBook Pro, and Mac mini. They join the rest of the Mac product line to form the strongest Mac lineup ever. This is the beginning of a transition to a new family of chips designed specifically for the Mac. The transition to Apple silicon will take about two years to complete, and these three systems are an amazing first step.

View at TechPowerUp Main Site

 

[xkm1948]

Whole bunch of buzzwords.

To me now macbook is dead. Apple is very clear in their intention to turn their walled garden into a walled prison.

 

[dicktracy]

The faster we move to ARM, the better. Just waiting for someone to do the same thing on the Windows side because Apple sure as hell ain't gonna sell their chips to anyone else.

 

[repman244]

The faster we move to ARM, the better. Just waiting for someone to do the same thing on the Windows side because Apple sure as hell ain't gonna sell their chips to anyone else.

And what exactly is better in ARM vs what we are currently using?

 

[pat-roner]

And what exactly is better in ARM vs what we are currently using?

I guess we'll know more about performance when the macs hits the real world, but perf/watt. More competition in chip makers

 

[TheoneandonlyMrK]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

 

[Ravenas]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

Maybe they collaborated with Trump.

 

[yotano211]

Maybe they collaborated with Trump.

there is no evidence of that

 

[xkm1948]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

I mean the target audience just need to hear apple and they will automatically open their wallet up, so why would they need to back up their claim?

 

[Ruru]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

Fastest iGPU... Ryzen 5 3400G's iGPU isn't that fast after all.

 

[dyonoctis]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

Apple has been very carefull with their wording, and they made really specific, but at the same time vague comparison : "faster than the best selling laptop/desktop pc, faster than any laptop with the same weight" without details, but I doubt that those are high end, and are most likely entry level computers used by people that are not doing professional/heavy work. They don't lie, but at the same time they don't really tell you anything.

We'll have to wait for the second batch (imac, macbook pro 16) to have a better idea of what Apple can offer for heavier task. If apple silicon can beat a decent x86 (like the 5800x/5900x, 4800h) on something like 3D rendering, you bet that they are going to brag about it instead of using vague comparison.

 

[xorbe]

The faster we move to ARM, the better.

ARM is also a walled prison, for the design teams. But I agree that someone needs to slash x86 legacy from x86_64 and start over there. That's a Hard Problem though (on the software and server side).

 

[TheoneandonlyMrK]

Fastest iGPU... Ryzen 5 3400G's iGPU isn't that fast after all.

Iris Xen and rDNA are it's competition not Vega and I prefer my boasts With benches personally.

 

[R0H1T]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

They did add the part about low power, all the while comparing it against Intel ICL(?) IIRC ~

It features the world's fastest CPU core in low-power silicon, the world's best CPU performance per watt

Which sounds reasonable enough given Axx SoC have been class leading chips in phones/tablets for quite a while & still are IIRC. For the tasks you're going to perform on a Mac, the optimization achieved with the M1 will be unparalleled & while in terms of raw performance AMD probably leads overall even they can't do the kind of integration Apple is doing here. Which is to say that the M1 is the best chips to do Mac things on a Mac, shocking I know

 

[zony249]

Not even the slightest bit of evidence for the fastest CPU and the fastest GPU.
Worse than an intel pr release.

I agree its very asshole marketing , touting "more powerful than 98% of laptops... etc". That said, their claims of the fastest CPU cores and iGPU are not wrong... Last year Anandtech covered the A13 Bionic, and concluded that the A13 performance core was as fast as Skylake and zen2, albiet the latter are clocked much faster (~4.7 - 5 GHZ) vs 2.66 GHz from A13. This year A14 improves that by about ~20%, but is also facing Zen3 and Willow Cove with roughly 20 ~ 25% increases each. Do keep in mind that that is A14, not M1. M1 does not have the same thermal confines as A14, so higher clocks are possible.

In terms of iGPU, the M1 does NOT have the fastest iGPU... Those go to the Xbox Series X... That said, comparing to Intel Iris Xe MAX, it is roughly 3% faster in pixel and texture fill rate, and wipes the floor with vega 11. overall it is faster than an RX 560. (I compared the texture fill rates and pixel fill rates, using their claims of 82 GTexels/s and 41GPixels/s for comparison)

Overall, is it impressive? yes.
Am I disappointed in the marketing? YES!
Is it a game changer? We will leave it to the developers and see what they say.

 

[Vya Domus]

What Apple isn't telling people is that they use a lot of dedicated hardware blocks to get this sort of performance, so whenever you hear photography this, video that, a million times faster, it's because of that not because of their amazing GPUs and CPUs. Of course that works because most of their users get macs specifically for those few programs that can be optimized that way, none of this would work in a Windows machine. So don't hold your breath for the ARM takeover in PCs, it ain't happening.

ARM is also a walled prison, for the design teams.

You can do whatever the hell you want with a custom ARM license, it will be costly sure, but it can be done.

 

[tabascosauz]

CPU side was impressive, but we all knew this was coming. Especially with the release of Hurricane, I don't think most people appreciate just how wide Apple cores have gotten since Swift. And now they have ever-increasing ML integration on their side as well.

Intel went after a similar design philosophy with Ice Lake and wants to complete the transition with Alder Lake big.LITTLE, but 10nm turned them their plans into soggy cardboard. Maybe Sunny Cove is a little underappreciated in hindsight thanks to Kaby and Coffee Lake clocking so high and 10+ being abysmal.

But hey, the low clocks work out perfectly for Apple, the company that had no heatpipe connecting the MBA CPU block and the fin stack...now that it's fanless, they can get rid of the entire cooler!

In all seriousness, there's a lot of talk about how Apple might just dial up the clocks and beat everyone. Something tells me that even with the help of N5 and N3 TSMC, the size of their core design might make it easier said than done...they don't just get a free pass on thermal density. That said, the way TSMC processes are going, they don't need to dial up the clocks if the cores just get keep getting wider - they sit right in the sweet spot.

 

[craigo]

 

[Fouquin]

To me now macbook is dead. Apple is very clear in their intention to turn their walled garden into a walled prison.

They're offering native translation, backwards compatibility, development support, and cross platform support. But sure, walled garden. It's a hell of a short wall.

And what exactly is better in ARM vs what we are currently using?

Well power efficiency for one. Unless you don't want more performance in lighter laptops, or you don't want 17+ hours of battery life. Some people (a lot of people) do though.

Core density. An ARM SoC can pack a lot into a small package thanks in part to core design. So more denser cores means more can fit in package like faster DRAM (so like 500GB/s instead of ~51) or fixed function logic blocks, or tensor blocks, or general purpose graphics blocks. All in much smaller space with a lot of the usual headaches like latency largely resolved.

Of course that works because most of their users get macs specifically for those few programs that can be optimized that way, none of this would work in a Windows machine. So don't hold your breath for the ARM takeover in PCs, it ain't happening

Let's see how this comment ages by the end of next week. Yes I have a crystal ball and it's pointed directly at Microsoft.

 

[Vya Domus]

Let's see how this comment ages by the end of next week. Yes I have a crystal ball and it's pointed directly at Microsoft.

Care to elaborate and why it is related with what I said ?

 

[Fouquin]

Care to elaborate and why it is related with what I said ?

Because Microsoft makes Windows? I'm being intentionally cryptic a bit here but I thought it was still fairly obvious what I was pointing to. This won't work on Windows unless Microsoft makes it work, so look toward Microsoft and come back next week to see how things might have changed.

 

[Vya Domus]

What hardly one payed attention to throughout these years is what kind of transistor budgets Apple is using to make these things, had they did that one would be able to see that their cores will soon become impossible to scale up. See, AMD, Intel and ARM will eventually add 200KB+ instruction/data caches and all of these other things Apple is using to boost performance for those lovely Geekbench scores, problem is Intel and AMD are extracting similar or more performance from designs that are way smaller and narrower already. Same with GPUs, there is so much you can fit on an SoC.

To sum it up, the efficiency of their use of transistor budgets is atrocious, "16 billion transistors" is not an achievement it's the hallmark of bad design with no foresight. They even famously said a long time ago that "it's easy to make something bigger, everyone can do that". Indeed, everyone can do it and when they will it's not going to be pretty.

Because Microsoft makes Windows? I'm being intentionally cryptic a bit here but I thought it was still fairly obvious what I was pointing to. This won't work on Windows unless Microsoft makes it work, so look toward Microsoft and come back next week to see how things might have changed.

You are not being cryptic, what you wrote genuinely makes no sense. What the hell is happening next week ?

 

[TheoneandonlyMrK]

I agree its very asshole marketing , touting "more powerful than 98% of laptops... etc". That said, their claims of the fastest CPU cores and iGPU are not wrong... Last year Anandtech covered the A13 Bionic, and concluded that the A13 performance core was as fast as Skylake and zen2, albiet the latter are clocked much faster (~4.7 - 5 GHZ) vs 2.66 GHz from A13. This year A14 improves that by about ~20%, but is also facing Zen3 and Willow Cove with roughly 20 ~ 25% increases each. Do keep in mind that that is A14, not M1. M1 does not have the same thermal confines as A14, so higher clocks are possible.

In terms of iGPU, the M1 does NOT have the fastest iGPU... Those go to the Xbox Series X... That said, comparing to Intel Iris Xe MAX, it is roughly 3% faster in pixel and texture fill rate, and wipes the floor with vega 11. overall it is faster than an RX 560. (I compared the texture fill rates and pixel fill rates, using their claims of 82 GTexels/s and 41GPixels/s for comparison)

Overall, is it impressive? yes.
Am I disappointed in the marketing? YES!
Is it a game changer? We will leave it to the developers and see what they say.

Without something to back their claims it's all talk, should be some actual testing done eventually.
It's not That impressive, many chips due out are suggested to be the next best thing.
Benches or don't let the door hit you ontt way out.

 

[crazyeyesreaper]

Fastest iGPU... Ryzen 5 3400G's iGPU isn't that fast after all.

They quote TFLOPs which is a terrible metric for performance (AMD usually has higher TFLOP ratings than NVIDIA still usually end up behind performance wise, they also do not state if its FP16 FP32 or FP64, if its FP16 than well AMD's Vega 11 has it beat handily in that useless metric at 3.492 TFLOPs. There is also the AMD Radeon RX Vega M GL in the Kabylake SoC which is performance wise nearly double the Vega 11 but has a TFLOP rating of 2.588. There is also the AMD Radeon RX Vega M GH which is rated at 3.656 TFLOPs and it has performance roughly similar to a Radeon 7970 / GTX 960 etc etc. So from a performance per watt example Apple may indeed be correct, but its likely not the fastest IGP.

As for pixel fill rate etc, that could determine performance but they also dont specify the memory arrangement or bandwidth. For example take a VEGA IGP add in some insane memory and performance scales quite noticeably. So its really just bs marketing but thats how the industry works I guess. Regardless its on ARM anyway so wont be much in the way of games to play on it lol.

 

[Fouquin]

What hardly one payed attention to throughout these years is what kind of transistor budgets Apple is using to make these things, had they did that one would be able to see that their cores will soon become impossible to scale up.

They're still 22% short of what TSMC can actually print reliably. Also it's exceptionally bold to assume core performance is actually what matters here. Cores have been strong for a decade and incremental improvements aren't hard to filter out every year. It's moving data around in a reasonable time frame that's been the biggest bottleneck. A proper high bandwidth, low latency fabric to seat all those shiny logic blocks on is going to do more for your performance than any transistor level tweaking or scaling. You need multiple terabytes per second end to end to make any of these CPU cores happy.

You are not being cryptic, what you wrote genuinely makes no sense. What the hell is happening next week ?

Wait. For. The. Microsoft. Announcement. About. ARM. And. Windows.

I hope that's plain enough.