Apple M1 chip: how to open an era
2020-12-06 08:01:02 【osc_ vzt0zpwr】
source | Old stone talks about the core
11 month 11 Japan , This year's trilogy has finally come to an end . The end of the play , Released Apple's own chip Apple Silicon The first generation of products ：M1 chip . The first flight M1 Chip Mac There are three types of machines ：MacBook Air、13 " MacBook Pro and MacMini. Regardless of the performance improvements announced at the launch , Or in recent days, the actual evaluation of these machines has come out ,M1 The performance and power consumption of the chip are beyond people's expectation , use “ subversion ” It's not too much to describe .
Today's article is about Apple M1 chip . I want to look at chip design , Let's see M1 Three main reasons why chips are so powerful .
Chip design PPA Optimize
When designing chips , One of the most important principles is to PPA The optimization of the , That is to optimize the power consumption of the chip as much as possible （Power）、 performance （Performance） And area （Area）. Usually , These three points can't be combined .
such as , In order to improve the performance of the chip , We can add multi-stage assembly lines 、 Increase the bus width 、 Or add a variety of hard core processing units , But at this point, it is likely to pay higher power consumption 、 And larger chip area at the cost of . conversely , If we want to design low-power chips , It's also possible to sacrifice some of the chip's performance . So in actual engineering practice , Power waste 、 Performance and area are often compromised 、 A balanced relationship . For a chip , There is no perfect design , Only a perfect balance . It's like giving you a limited budget to buy ingredients for cooking , It's hard to buy fish , Buy Bear's paws again .
But this time Apple M1 The release of the chip , The most intuitive feeling is , With the same budget , You can fish and bear's paws , Both .
Apple released this picture at the press conference , It compares M1 Chip and some laptop CPU, But Apple didn't say who . For the convenience of narration , Let's call it here “ Three chips ”. In this picture , The horizontal axis is power consumption , The vertical axis is performance . You can see , For a given power consumption standard ,M1 Its performance is twice that of Zhang San . For a given performance requirement ,M1 The power consumption of Zhang San is a quarter of that of Zhang San .
in other words , In the two dimensions of high performance and low power consumption, which tend to go up and down ,M1 Chips have also been greatly improved . except CPU outside ,M1 Inside GPU Similar performance improvements and power reductions have been achieved . Apple says ,M1 With the best in the world right now CPU Per watt performance , And the world's fastest integrated graphics card .
No matter what language marketing uses , Running points is always a factor that can't be bypassed . Although Apple officials didn't say anything about running points , But all kinds of tests have already given quite high scores .
from Geekbench From the running points of , This time it's on board M1 Three products of the chip , Whether it's single or multi-core , They've got quite a high score ：
Mac Mini: 1682 / 7097
MacBook Air: 1687 / 7433
MacBook Pro: 1714 / 6802
in fact , They've all got more than one core AMD Just released the Reebok 9 5950X, It also surpasses Intel's TigerLake Flagship Edition , That is to say 11 Generation of core i7-1165G7.
Besides , The single core and multi-core scores of these three products also exceed those currently on sale 2019 paragraph 16 " MacBook Pro. in other words , Single from CPU From the running points of , This new style macbook air Faster than Apple's flagship laptop just released last year .
It's not just performance , carrying M1 The battery life of the chip's notebook has also been greatly improved , The highest can be achieved 20 Hours .
in my opinion ,M1 The chip has achieved such a huge energy efficiency improvement , There are three main reasons ： One is the use of 5 Nanotechnology , The second is some chip architecture innovation , The third is the deep optimization of hardware and software , Let's talk about it one by one .
TSMC 5 Nanotechnology
The manufacturing process , Many articles don't mention , Or not paying attention to . in fact , Advanced technology for chip energy efficiency improvement , It plays a very important role . Because of the improvements in chip performance over the past decade , There are more than 60% Benefit directly or indirectly from the improvement of semiconductor technology , And only 17% From the upgrade of chip architecture .AMD can “ Overtaking in curve ” One of the main reasons for Intel , It's abandoning Grosvenor , The embrace of TSMC .
M1 It should be the first in the world at present 、 And the only one that uses TSMC 5 Nano technology notebook processor chip , Which includes 160 Billion transistors . According to TSMC's data , And the previous generation of 7 Compared with nanotechnology , Use 5 Transistors made by nanotechnology ： Density increases 80%, Speed up 15%, Reduced power consumption 30%. With a new manufacturing process , You can keep the chip area constant , Put more transistors into a chip , And the power consumption of these transistors is lower 、 Higher performance .
TSMC process roadmap （ The picture is from WikiChip）
in other words , Even if Apple does nothing , Simply put A13 For chip 5 Nanotechnology comes out , Ideally, you can achieve this “ free ” Performance of 、 Power and area optimization . Of course, the actual situation is much more complicated than this .
Need to know ,2019 paragraph 16 " MacBook Pro It's still the ninth generation of Intel CPU, Using Intel 14 Nanotechnology , This is the same as TSMC's 5 Nanotechnology is at least two generations old . So it's not hard to understand why in terms of running points , carrying M1 Of MacBook Air It will reduce dimension and strike 16 " MacBook Pro.
Apple M1 Package diagram （ The picture is from Apple ）
Chip architecture innovation
In particular, Apple's UMA structure , That's the unified memory architecture shown this time . One of the big principles of chip design is , The closer the data is stored to where it is used , The higher the performance 、 The lower the power consumption . So Apple took the memory particles that were originally on the circuit board , Integrated into the chip package . The main advantage of this is to make the chips CPU、GPU、AI The engine can access memory faster , At the same time, it greatly reduces the power consumption of data transmission . Besides , Each module can share memory , It also saves a lot of data handling 、 The cost of copying .
Apple UMA Structural sketch （ The picture is from Apple ）
It is worth noting that , This kind of architecture design and encapsulation method is not unique to apple , Actually in NVIDIA A100GPU、AMD Of Rome processor , And the high end of Intel and Xilinx FPGA In the chip , They all use similar methods , Multiple different computing and memory units can be integrated in the same chip package .
There are many ways to implement this kind of encapsulation , such as AMD The pellets used chiplets, And Intel's EMIB technology , And salings is FPGA Used in SSI, That is, stacked silicon interconnection technology and so on . The industry calls these technologies 2.5D Packaging technology , That is to connect multiple small silicon wafers in the horizontal direction , And then make a big chip .
EMIB Structural sketch （ Picture from Intel ）
Here's one more thing , Intel also has a name Foveros Of 3D Packaging technology .Foveros From Greek , The original intention is “ frigging awesome ”. This technology doesn't scale horizontally , It's vertical expansion . in other words , It can granulate memory 、CPU、GPU、 There are other chip units , Fold them together like sandwiches , This makes memory and CPU The distance is reduced to 0.1 Mm or so , So it will further increase the memory bandwidth 、 Reduce transmission delay , And it doesn't increase the chip area . At Intel Lakefiled CPU That's what we're using here .
Foveros 3D Package diagram （ Picture from Intel ）
Collaborative optimization of Apple Ecology
Apple M1 The third reason for the significant improvement in chip energy efficiency , It's the collaborative deep optimization of Apple's hardware and software , This, I think, is Apple's biggest competitive advantage . This is because , As mentioned above 5 Nanotechnology , Or chip architecture and packaging technology upgrades , Other manufacturers may also （ Or it will ） Master and adopt these technologies . But only by combining hardware and software 、 Deep optimization of operating system and ecology , It's Apple's unique . And that's why there are, and only apple can, based on arm Architecturally CPU The reason why it's really made a great product to sell .
Some people say , Apple M1 The advent of the chip proves that Wintel The alliance is not hard to break , I don't agree with that . be based on arm The consumption level of Architecture CPU It has been said for many, many years , But if you look at the current hardware and software manufacturers other than apple, who can take this responsibility ？ Intel amd I'm sure I won't do something like this , Needless to say , Microsoft has tried to do something based on arm The ecology of , But it ended in failure . So let Lenovo 、 Dale does it arm The notebook , Let's not let the high pass NVIDIA cpu Manufacturers are going to do ecology , In fact, it is unrealistic at present .
As for software and hardware co optimization, I quite agree with a big v The dragon's reply , You can go and have a look . as long as pc The hardware and software manufacturers of our company are still fighting for each other , As long as different software and hardware still need to be compatible with each other and adapt , Whether it's the operating system 、 development tool 、 drive 、CPU、GPU、 Memory, etc , This ecosystem is hard to crack . Sometimes the ecosystem is like a snowball falling from the top of a mountain , Once you're involved in him, it's hard to escape on your own , Let alone change its trajectory .
Apple OS And ecology （ The picture is from Apple ）
by comparison , Apple has its own ecology , This time, M1 The release of the chip , It also officially completed the ecological closed loop . You can say that Apple ecology is too closed , Or problems like this , But only this closed ecology can produce something like M1 Chips and things like that .
from Anandtech You can see it in this picture , In the past 5 In the year , Intel chip's single threaded performance has improved 28%, Apple is up nearly 3 times . therefore , Apple uses self-developed Apple Silicon chip , It's also an inevitable choice .
（ The picture is from AnadTech）
however , Intel AMD The difference is , You can't buy this alone M1 chip . In order to use M1 Energy efficiency of , You have to buy the whole ecology of apples . And even with closed-loop ecology like apple , Full turn arm Architecture also takes two years , This involves a lot of ecological migration , It will also create a lot of extra work pressure for developers . So this ecological transformation is not easy .
Apple M1 The release of the chip , It's also based on arm The architecture of the consumer class processor is back on the stage of history , And start a frontal attack based on x86 Architecture of traditional processors . Although at present x86 The architecture is arm It's too early to replace , But in many areas , such as mac Computer is good at video and image creation ,M1 Chips have officially overturned people's perception of energy efficiency . I believe that with the emergence of more follow-up products , There will be more subversion . meanwhile ,x86 How the camp responds , It's also a very important point to pay attention to .
- C++ 数字、string和char*的转换
- Won the CKA + CKS certificate with the highest gold content in kubernetes in 31 days!
- C + + number, string and char * conversion
- C + + Learning -- capacity() and resize() in C + +
- C + + Learning -- about code performance optimization
C + + programming experience (6): using C + + style type conversion
Latest party and government work report ppt - Park ppt
Online ID number extraction birthday tool
Field pointer? Dangling pointer? This article will help you understand!
GVRP of hcna Routing & Switching
- LeetCode 91. 解码方法
- Seq2seq implements chat robot
- [chat robot] principle of seq2seq model
- Leetcode 91. Decoding method
- HCNA Routing＆Switching之GVRP
- GVRP of hcna Routing & Switching
- HDU7016 Random Walk 2
- [Code+＃1]Yazid 的新生舞会
- CF1548C The Three Little Pigs
- HDU7033 Typing Contest
- HDU7016 Random Walk 2
- [code + 1] Yazid's freshman ball
- CF1548C The Three Little Pigs
- HDU7033 Typing Contest
- Qt Creator 自动补齐变慢的解决
- HALCON 20.11：如何处理标定助手品质问题
- HALCON 20.11：标定助手使用注意事项
- Solution of QT creator's automatic replenishment slowing down
- Halcon 20.11: how to deal with the quality problem of calibration assistant
- Halcon 20.11: precautions for use of calibration assistant
- "Top ten scientific and technological issues" announced| Young scientists 50 ² forum
- Reverse linked list
- JS data type
- Remember the bug encountered in reading and writing a file
- Singleton mode
- 在这个 N 多编程语言争霸的世界，C++ 究竟还有没有未来？
- In this world of N programming languages, is there a future for C + +?
- js Promise
- js 数组方法 回顾
- ES6 template characters
- js Promise
- JS array method review
- 【Golang】️走进 Go 语言️ 第一课 Hello World
- [golang] go into go language lesson 1 Hello World