Daily Productive Sharing 075 - Apple 近十年来最大的创新突破

One helpful tip per day:)

(The English version follows)

最近 Apple 发布了基于 M1 芯片的 Mac,在我看来,这可能是 Apple 近十年来最重要的一次革新。这款 Mac 诞生的意义可能不亚于 iPhone。为什么这么说呢?我们陆续整理了近期几篇深度的分析报道,全部放到了今天的分享里。

  1. M1 芯片整合了 CPU,GPU,内存和 NPU,因此数据在不同组件转移的距离小得多,完全颠覆了传统电脑的架构设计;

  2. 因为上述原因,所以 M1 的功耗非常小,不需要额外的散热系统,大大延长了电池寿命;

  3. 因为 GPU,CPU 和 NPU 共用内存,所以调度起来更方便,因而内存的使用效率更高,也就不需要传统架构那么多的内存;

  4. iPhone,iPad 和 Apple watch 都是软硬件结合,Apple 一手掌控了两者;而在 M1 诞生之前,Mac 是 Apple 唯一不掌控两者的产品。M1 的诞生意味着 Apple 掌控了 Mac 的软硬件,至此,拼图的最后一块也补上了;

  5. 同时掌控软硬件,意味着 Apple 对 Mac 有更强的话语权,进一步加速革新;

  6. 传统的 CPU 厂商可能不再是 Apple 在这一领域的直接竞争对手,毕竟他们只专注在 CPU 这一单一领域。最有威胁的竞争对手反倒是高通,毕竟它也专注在整合芯片的研发上。

  7. 如果按照上面这条逻辑,华为的海思倒也是潜在竞争对手。如果海思能够在当下浴火重生,很有可能生产自研的整合芯片;

  8. 当下的软件和之前的软件已经大不一样。之前的软件只会按照设计时的逻辑运作,而有了机器学习加持的软件会“学习”用户的使用习惯,自我“进化”来适应。这样的需求意味着硬件上更高的要求,所以 NPU 是 M1 里的重要一环。传统的硬件设计里很少有类似的专有硬件(Intel 有 Modivious 计算棒,但是没有普及开来);

  9. CPU 的时钟频率可能不再重要,因为这并不是用户最关心的。用户只关心一台电脑能处理多少任务,能够有多少续航。传统 CPU 强调这些,是因为这是最容易量化的指标。这倒是让我想起来为什么很多人去健身房最喜欢举铁,因为这也是一个极其容易量化的运动。

“We are giddy”—interviewing Apple about its Mac silicon revolution

To hear Apple's Craig Federighi tell the story, it sounds a bit like a callback to Steve Wozniak in a Silicon Valley garage so many years ago.

And for a long time, Intel's performance was top-notch. But in recent years, Intel's CPU roadmap has been less reliable, both in terms of performance gains and consistency.

The M1 is essentially a superset, if you want to think of it relative to A14. Because as we set out to build a Mac chip, there were many differences from what we otherwise would have had in a corresponding, say, A14X or something.

So we've been building these great technologies over a decade, and then several years back, we said, "Now it's time to use what we call the scalable architecture." Because we had the foundation of these great IPs, and the architecture is scalable with UMA.

Then we said, "Now it's time to go build a custom chip for the Mac," which is M1. It's not like some iPhone chip that is on steroids. It's a whole different custom chip, but we do use the foundation of many of these great IPs.

UMA essentially means that all the components—a central processor (CPU), a graphics processor (GPU), a neural processor (NPU), an image signal processor (ISP), and so on—share one pool of very fast memory, positioned very close to all of them.

We not only got the great advantage of just the raw performance of our GPU, but just as important was the fact that with the unified memory architecture, we weren't moving data constantly back and forth and changing formats that slowed it down. And we got a huge increase in performance.

Where old-school GPUs would basically operate on the entire frame at once, we operate on tiles that we can move into extremely fast on-chip memory, and then perform a huge sequence of operations with all the different execution units on that tile. It's incredibly bandwidth-efficient in a way that these discrete GPUs are not. And then you just combine that with the massive width of our pipeline to RAM and the other efficiencies of the chip, and it’s a better architecture.

In other words, the MacBook Air and Pro should perform similarly when bursts of speed are needed, but the Pro can maintain that performance for longer—critical for many demanding workflows like video editing. In such scenarios, the Air would throttle sooner to stay cool. According to Apple, the M1 scales to better performance the more power you can give it.

Steve Jobs’s last gambit: Apple’s M1 Chip

And my excitement about the new Apple Silicon is not tied to a single chip, a single computer, or a single company. It is really about the continuing — and even accelerating — shift to the next phase of computing.

Today’s computers are shape-shifting — they are servers in the clouds, laptops in our bags, and phones in our pockets. The power of a desktop from just five years ago is now packed inside a keyboard and costs a mere $50-a-pop from Raspberry Pi. Cars and TVs are as much computers as they are anything else.

Jobs learned the hard way that, to stay competitive, Apple had to make and control everything: the software, the hardware, the user experience, and the chips that power it all.

Much of it can be summed up in this line from a 2017 piece: “Don’t depend on a third party to be an enabler of your key innovations and capabilities.”

“Steve used to say that we make the whole widget,” Joswiak told me. “We’ve been making the whole widget for all of our products, from the iPhone, to the iPads, to the watch. This was the final element to making the whole widget on the Mac.”

Apple’s chip, hardware, and software teams work together to define the future systems to integrate them tightly.

Machine learning will define the capabilities of the software in the future.

As a result, the entire chip can access data without copying it between different components and going through interconnects. This allows them to access memory with very low latency and at a higher bandwidth. The result is a much better performance with less power consumption.

I don’t think AMD and Intel are Apple’s competitors. We should be looking at Qualcomm as the next significant laptop chip supplier. Apple’s M1 is going to spark an interest in new architectures from its rivals.

This approach to integration into a single chip, maximum throughput, rapid access to memory, optimal computing performance based on the task, and adaptation to machine learning algorithms is the future — not only for mobile chips, but also for the desktop and laptop computer. And this is a big shift, both for Apple and for the PC industry.

The most significant of these shifts came in 2005. Hampered by the fading Power PC ecosystem, the company made a tough decision to switch to the superior Intel ecosystem. The shift to x86 architecture came alongside a new operating system — the Mac OS X. The change caused a lot of disruption, both for developers and the end customers.

In his most recent keynote, Apple CEO Tim Cook pointed out that one in two new computers sold by Apple is being bought by the first time Mac buyers.

The idea that bigger numbers are a proxy for better quality has become ingrained in modern thinking about laptops and desktops. This mental model will be a big challenge for Apple.

As he sees it, just as no one cares about the clock speed of the chip inside an iPhone, the same will be true for the new Macs of the future. Rather, it will all be about how “many tasks you can finish on a single battery life.”

Today, our apps are becoming more personal and smarter as we use them. Our interactions define their capabilities. It is always learning.

As Apple merges the desktop, tablet, and phone operating systems into a different kind of layer supported by a singular chip architecture across its entire product line-up, traditional metrics of performance aren’t going to cut it.

The M1 — at least, in theory — uses the UMA to eliminate all that need to move the data back and forth. On top of that, Apple has a new optimized approach to rendering, which involves rendering multiple tiles in parallel and has allowed the company to remove complexity around the video systems.

For example, machine learning is going to play a bigger role in our future, and that is why neural engines need to evolve and keep up. Apple has its algorithms, and it needs to grow its hardware to keep up with those algorithms.

With M1 Macs, memory isn't what it used to be

The M1 processor’s memory is a single pool that’s accessible by any portion of the processor. If the system needs more memory for graphics, it can allocate that. If it needs more memory for the Neural Engine, likewise. Even better, because all the aspects of the processor can access all of the system memory, there’s no performance hit when the graphics cores need to access something that was previously being accessed by a processor core.

The unified memory architecture in the M1 is one of the reasons these Macs are so amazingly fast—but all Mac users are going to have to relinquish some of our assumptions about how our computers work, and how they’re configured.

The Mac's Graduation

Approximately 90% of Apple users don’t use, and probably never will use, a Mac.

However, 2020 was a record fiscal year for the Mac in terms of revenue and the number of new users was near a record high.

It is telling that Apple didn’t see the need to change the MacBook Air’s design despite the fact that it is being powered by Apple Silicon. This is evidence of the Apple Silicon transition being akin to the Mac graduating and entering a new phase in life.

The transition would not only give Apple the kind of control over the Mac that it yearned for, but more importantly, Apple Silicon would open new doors to push the Mac forward in ways that simply weren’t possible with Intel.

There are 7x more people using iPhones than Macs. There are 2x more people using iPads than Macs. Some think that Apple Silicon will dramatically change these ratios by increasing the Mac’s addressable market. Caution is needed in running too far with such thinking.

Even though Mac portables and iPads may handle many similar workflows, that doesn’t mean that both devices should lose their core identity.

Share

如果你想更好地管理时间,并且减轻自己的压力,不妨试试 BRNR List

如果你也想成为更高效的人,欢迎加入我们的 TG group

也欢迎订阅我们的 TG channel


Apple recently released a Mac based on the M1 chip, which, in my opinion, may be the most important innovation from Apple in a decade. This Mac may be as significant as the first generation iPhone. We have compiled several recent in-depth analysis reports, all of which are included in today's post.

  1. The M1 chip integrates CPU, GPU, memory, and NPU, so that data can be transferred over much smaller distances between components, completely disrupting the architectural design of traditional computers.

  2. Because of the above, the M1 consumes very little power and does not require an additional cooling system, which greatly extends its battery life.

  3. Because the GPU, CPU and NPU share memory, they are easier to coordinate and thus use memory more efficiently and do not require as much memory as traditional architectures.

  4. iPhone, iPad, and Apple watch are a combination of hardware and software, and Apple controls them all; until M1, the Mac was the only product Apple didn't control both sides. M1 meant that Apple could control the Mac's hardware and software, and thus the last piece of the puzzle was completed.

  5. Controlling both hardware and software means Apple has a stronger voice over the Mac, further accelerating innovation.

  6. The traditional CPU vendors may no longer be direct competitors to Apple in this area, as they focus on CPU alone. The most threatening competitor would be Qualcomm, which also focuses on integrated chip development.

  7. If we follow the above logic, Huawei's Hisilicon is also a potential competitor. If Hisilicon can survive, it is likely to produce its own integrated chips.

  8. The current generation of software is very different from the previous generation. Whereas software only works according to the logic it was designed for, software with machine learning "learns" the user's habits and "evolves" to adapt. This demand means higher hardware requirements, so NPU is an important part of M1. There is little similar proprietary hardware in traditional hardware designs (Intel has the Modivious computing stick, but it is not widely available).

  9. The clock frequency of the CPU may no longer be important, because it is not the most important thing for users. The only thing that users care about is how many tasks a computer can handle and how much battery life it can have. Traditional CPUs emphasize this because it's the easiest metric to quantify. This reminds me why many people like to go to the gym and lift iron, because it's also an extremely easy exercise to quantify.

“We are giddy”—interviewing Apple about its Mac silicon revolution

To hear Apple's Craig Federighi tell the story, it sounds a bit like a callback to Steve Wozniak in a Silicon Valley garage so many years ago.

And for a long time, Intel's performance was top-notch. But in recent years, Intel's CPU roadmap has been less reliable, both in terms of performance gains and consistency.

The M1 is essentially a superset, if you want to think of it relative to A14. Because as we set out to build a Mac chip, there were many differences from what we otherwise would have had in a corresponding, say, A14X or something.

So we've been building these great technologies over a decade, and then several years back, we said, "Now it's time to use what we call the scalable architecture." Because we had the foundation of these great IPs, and the architecture is scalable with UMA.

Then we said, "Now it's time to go build a custom chip for the Mac," which is M1. It's not like some iPhone chip that is on steroids. It's a whole different custom chip, but we do use the foundation of many of these great IPs.

UMA essentially means that all the components—a central processor (CPU), a graphics processor (GPU), a neural processor (NPU), an image signal processor (ISP), and so on—share one pool of very fast memory, positioned very close to all of them.

We not only got the great advantage of just the raw performance of our GPU, but just as important was the fact that with the unified memory architecture, we weren't moving data constantly back and forth and changing formats that slowed it down. And we got a huge increase in performance.

Where old-school GPUs would basically operate on the entire frame at once, we operate on tiles that we can move into extremely fast on-chip memory, and then perform a huge sequence of operations with all the different execution units on that tile. It's incredibly bandwidth-efficient in a way that these discrete GPUs are not. And then you just combine that with the massive width of our pipeline to RAM and the other efficiencies of the chip, and it’s a better architecture.

In other words, the MacBook Air and Pro should perform similarly when bursts of speed are needed, but the Pro can maintain that performance for longer—critical for many demanding workflows like video editing. In such scenarios, the Air would throttle sooner to stay cool. According to Apple, the M1 scales to better performance the more power you can give it.

Steve Jobs’s last gambit: Apple’s M1 Chip

And my excitement about the new Apple Silicon is not tied to a single chip, a single computer, or a single company. It is really about the continuing — and even accelerating — shift to the next phase of computing.

Today’s computers are shape-shifting — they are servers in the clouds, laptops in our bags, and phones in our pockets. The power of a desktop from just five years ago is now packed inside a keyboard and costs a mere $50-a-pop from Raspberry Pi. Cars and TVs are as much computers as they are anything else.

Jobs learned the hard way that, to stay competitive, Apple had to make and control everything: the software, the hardware, the user experience, and the chips that power it all.

Much of it can be summed up in this line from a 2017 piece: “Don’t depend on a third party to be an enabler of your key innovations and capabilities.”

“Steve used to say that we make the whole widget,” Joswiak told me. “We’ve been making the whole widget for all of our products, from the iPhone, to the iPads, to the watch. This was the final element to making the whole widget on the Mac.”

Apple’s chip, hardware, and software teams work together to define the future systems to integrate them tightly.

Machine learning will define the capabilities of the software in the future.

As a result, the entire chip can access data without copying it between different components and going through interconnects. This allows them to access memory with very low latency and at a higher bandwidth. The result is a much better performance with less power consumption.

I don’t think AMD and Intel are Apple’s competitors. We should be looking at Qualcomm as the next significant laptop chip supplier. Apple’s M1 is going to spark an interest in new architectures from its rivals.

This approach to integration into a single chip, maximum throughput, rapid access to memory, optimal computing performance based on the task, and adaptation to machine learning algorithms is the future — not only for mobile chips, but also for the desktop and laptop computer. And this is a big shift, both for Apple and for the PC industry.

The most significant of these shifts came in 2005. Hampered by the fading Power PC ecosystem, the company made a tough decision to switch to the superior Intel ecosystem. The shift to x86 architecture came alongside a new operating system — the Mac OS X. The change caused a lot of disruption, both for developers and the end customers.

In his most recent keynote, Apple CEO Tim Cook pointed out that one in two new computers sold by Apple is being bought by the first time Mac buyers.

The idea that bigger numbers are a proxy for better quality has become ingrained in modern thinking about laptops and desktops. This mental model will be a big challenge for Apple.

As he sees it, just as no one cares about the clock speed of the chip inside an iPhone, the same will be true for the new Macs of the future. Rather, it will all be about how “many tasks you can finish on a single battery life.”

Today, our apps are becoming more personal and smarter as we use them. Our interactions define their capabilities. It is always learning.

As Apple merges the desktop, tablet, and phone operating systems into a different kind of layer supported by a singular chip architecture across its entire product line-up, traditional metrics of performance aren’t going to cut it.

The M1 — at least, in theory — uses the UMA to eliminate all that need to move the data back and forth. On top of that, Apple has a new optimized approach to rendering, which involves rendering multiple tiles in parallel and has allowed the company to remove complexity around the video systems.

For example, machine learning is going to play a bigger role in our future, and that is why neural engines need to evolve and keep up. Apple has its algorithms, and it needs to grow its hardware to keep up with those algorithms.

With M1 Macs, memory isn't what it used to be

The M1 processor’s memory is a single pool that’s accessible by any portion of the processor. If the system needs more memory for graphics, it can allocate that. If it needs more memory for the Neural Engine, likewise. Even better, because all the aspects of the processor can access all of the system memory, there’s no performance hit when the graphics cores need to access something that was previously being accessed by a processor core.

The unified memory architecture in the M1 is one of the reasons these Macs are so amazingly fast—but all Mac users are going to have to relinquish some of our assumptions about how our computers work, and how they’re configured.

The Mac's Graduation

Approximately 90% of Apple users don’t use, and probably never will use, a Mac.

However, 2020 was a record fiscal year for the Mac in terms of revenue and the number of new users was near a record high.

It is telling that Apple didn’t see the need to change the MacBook Air’s design despite the fact that it is being powered by Apple Silicon. This is evidence of the Apple Silicon transition being akin to the Mac graduating and entering a new phase in life.

The transition would not only give Apple the kind of control over the Mac that it yearned for, but more importantly, Apple Silicon would open new doors to push the Mac forward in ways that simply weren’t possible with Intel.

There are 7x more people using iPhones than Macs. There are 2x more people using iPads than Macs. Some think that Apple Silicon will dramatically change these ratios by increasing the Mac’s addressable market. Caution is needed in running too far with such thinking.

Even though Mac portables and iPads may handle many similar workflows, that doesn’t mean that both devices should lose their core identity.

Share

Try our sustainable productivity tool BRNR List