If you are a young, ambitious entrepreneur starting from scratch, and you have the opportunity to return to any period in the IT industry to start your business, which period would you choose? I believe you would choose the exciting 1970s.
If the 1950s belonged to the nascent semiconductor technology and the era of vacuum tube computers, and the 1960s was the dawn of the semiconductor industry and the rise of large commercial mainframes, then the 1970s heralded the dawn of the consumer market in the IT industry. Several companies that emerged during this time still dominate the IT world today.
In 1975, Paul Allen and Bill Gates began their entrepreneurial journey and founded Micro-Soft the following year, officially renamed Microsoft in 1978, which is today’s Microsoft.
On April 1, 1976, Steve Jobs and his childhood friend Steve Wozniak co-founded Apple Computer in the garage of Jobs’ parents’ house in California.
In 1979, Hermann Hauser, who had just graduated with a PhD from Cambridge University, and Chris Curry established Acorn Computers.
(Acorn Computer Ltd Logo)
While Microsoft’s and Apple’s current status is well-known, many people may not be familiar with Acorn Computers from the UK. If we say that Acorn Computers is the predecessor of ARM, which today provides system architecture for billions of mobile terminal chips, people might suddenly realize its significance.
We know that Apple initiated the personal computer era, but due to Jobs’ stubbornness, Apple lost to companies like IBM, Compaq, and Dell that followed the Wintel alliance route. Apple’s Mac remains a niche representative in the PC era. Acorn Computers also couldn’t survive its glorious days but left behind ARM as a cornerstone supporting the powerful computing power of Apple, Android, and other smart mobile terminals.
In June 2020, Apple announced at WWDC 2020 that it would equip its Mac products with self-developed chips featuring the ARM instruction set, completely realizing the integration of mobile and PC computing architectures. This choice by Apple seems to represent a perfect cycle. It is worth noting that ARM was co-founded by Acorn Computers, Apple, and a semiconductor foundry company, VLSI.
In this issue, we will return to the historical site of Acorn’s birth, revisiting the circumstances that led to ARM’s success and the mobile era opened by ARM’s processor architecture.
Innovators Emerge: The Birth of ARM at Acorn Computers
Back in the early 1970s, computers were still expensive devices used only by large corporations and government agencies, and inventing and building personal computers was the trend among electronic enthusiasts of the time.
When Jobs witnessed Wozniak’s talent for building computers, he seemed to have committed to the personal computer path. Finally, in 1976, they produced the prototype of the Apple I and successfully sold the first batch of Apple computers.
Initially, Jobs insisted on using the name Apple for the company, reportedly because he had spent two years in India living mainly on apples, which became his source of strength; and also because he hoped that Apple would rank high in the telephone directory.
This clever idea seems to have inspired Acorn’s founders, Hauser and Curry, who named their company Acorn in 1979, hoping that their personal computer business would grow into a thriving oak tree like an acorn. They also wished to rank ahead of Apple in the telephone directory.
Who would have thought that these two companies named after “fruits” would each become giants in the fields of smart mobile terminals and mobile terminal processors?
At that time, everything was just beginning. Whether it was the tech enthusiasts tinkering with electronic devices in garages or a physics PhD and a marketing genius without much technical background from Cambridge, they could all venture into the promising direction of personal computers, reflecting the vibrant and raw atmosphere at the dawn of an industry.
For Acorn, besides the two ambitious founders, two other figures became crucial. One was Acorn’s first technician, Sophie Wilson, who had just graduated from Cambridge University with a degree in mathematics. In simple terms, she was one of the earliest developers of Acorn Computers and the first developer of the ARM architecture instruction set. The other was the legendary engineer Steve Furber, who, along with Wilson, produced a prototype that satisfied the BBC in just a week. He then worked with Wilson on the development of the ARM processor, focusing on chip design.
Truly, the era creates geniuses, and geniuses shape the era.
The first generation of affordable personal computers launched by Acorn, the Acorn Atom, helped it initially penetrate the home market. The real breakthrough for Acorn in the microcomputer market came in 1981 when it secured a £1.3 million order from the BBC Micro. This project was made possible by Chris’s proactive efforts and Hermann’s successful negotiations internally, which inspired Wilson and Furber’s potential. Before the BBC team came for inspection again, the two used the still-in-development Acorn Proton version as a prototype, working tirelessly for five days and nights to create a physical prototype of the BBC Micro. Subsequently, the BBC Micro was a huge success, selling 1.5 million units and winning the Queen’s Award for Technology in 1984.
(Participants in the BBC Micro commemorative event in 2008)
The impetus for Acorn to develop its own processor arose after IBM launched the second generation of microcomputers for the commercial market in 1983, prompting Acorn to enter the microcomputer commercial market. However, the 2MHz 6502 processor from MOS Technology, used in the BBC Micro, was unable to meet the new hardware requirements, and there were no satisfactory processors available on the market for Acorn.
It is said that Wilson and Furber explored all possible chips at the time and found Intel’s 286 chip to be decent. However, when they approached Intel for collaboration to obtain a license for the 286 chip, they were flatly rejected. This refusal would cost Intel a nightmare competitor for the new century.
The balance of opportunity tilted in favor of Acorn. At that time, the Berkeley RISC Plan white paper proposed by Professor David Patterson at the University of California, Berkeley, inspired Wilson and Furber to begin developing a new generation of BBC microcomputers based on a RISC 32-bit microprocessor chip.
(Sophie Wilson introducing the ARM development process)
In October 1983, the ARM project was officially launched. Following the RISC architecture’s simplified approach, Wilson quickly wrote the first ARM prototype using BBC Basic. After 18 months of research and testing, in April 1985, Acorn’s chip foundry, VLSI, produced the first ARM chip using the RISC instruction set, with ARM standing for Acorn RISC Machine.
(ARM-1 chip on the BBC Micro)
On April 26, 1985, at 1 PM, the first batch of ARM chips returned from VLSI and were directly投入到开发系统中,并经过一两次调整后启动。在下午3点,屏幕中显示出:“ Hello World,I am Arm”。全球第一款商业 RISC 处理器——ARM-1在Acorn电脑公司成功运行。
不过,此时的Acorn电脑公司则因为经营失误而大幅亏损,Acorn不得不在1985年将自己的近一半股权以1200万英镑的低价转让给意大利的Olivetti电脑公司,用以偿还债务。在ARM芯片成功研制之后,Acorn以ARM开发系统作和一款Acorn Archimedes系统为商用产品进行销售。
此后,Acorn电脑的发展已经波澜不惊。因为在1990年,由Acorn电脑公司、苹果公司和那家代工厂VLSI合作创办的ARM(全称是Advanced RISC Machines Ltd)公司成立了。
新的历史进程和荣光将属于ARM。
RISC的星火:ARM选对了赛道
在继续ARM的故事前,我们有必要简单聊下指令集这个生僻的领域。
1961年,IBM在小沃森的支持下,由副总裁文森特·利尔森牵头,准备投入50亿美元进行IBM360电脑的开发。1964年,IBM 360系列电脑研制成功,成为划时代的产品。
而在研制过程中,IBM攻克了计算机指令集、集成电路、可兼容操作系统、数据库等一系列软硬件难关,也为此申请了300多项专利技术。其中,System-360系统,所集成的全新通用指令集架构(Universal Instruction Set Architecture),成为计算机发展史上第一种商用的指令集架构。
(IBM System360大型机)
可以说,指令集的本质,就是硬件和软件代码之间沟通的一套“标准语言”。处理器就是那个核心硬件,操作系统就是基础软件。电脑中运行的的软件就要基于指令集的架构而开发,才能够达成正常运行的效果。这一从处理器到操作系统,再到基础应用软件的搭建模式,便是以指令集架构为基础的一种“标准语言”的兼容生态。
我们常说的处理器架构,就是说以指令集架构为基础搭建起来的处理器。一旦指令集和操作系统组合绑定,将形成其他玩家难以逾越的“生态墙”。现在我们知道,引领PC时代最牢不可破的“生态墙”就是由英特尔x86处理器和微软Windows操作系统所构建的Wintel联盟。而移动互联时代的“生态墙”正是由ARM的RISC指令集架构形成的移动端处理器生态和Android、iOS操作系统构建起的ARM联盟。
指令集系统在发展过程中,分化出两个截然不同的优化方向:复杂指令系统计算(Complex Instruction Set Computing,CISC)和精简指令系统计算(Reduced Instruction Set Computing,RISC)。
CISC就是通过设置一些功能复杂的指令,把一些原来由软件实现的、常用的功能改用硬件的指令系统实现,以此来提高计算机的执行速度,其特点是设计复杂,功耗高,指令的执行时间不同,其优势在于高计算性能。而RISC的设计思路是尽量简化计算机指令功能,只保留那些功能简单、能在一个节拍内执行完成的指令,而把较复杂的功能用一段子程序来实现,其特点是每条指令执行时间相同,可以减少指令的平均执行周期,提高工作主频速度,另外一个优点就是功耗极低,但在通用性计算上面不及CISC。
(著名的英特尔8086处理器)
尽管在80年代,学术界认为CISC已经过时。但是由于英特尔在开发8086处理器时还没有RISC指令集,所以采用了CISC的设计,此后英特尔的处理器系列都采用了CISC指令集。
此后由于可观的收入,使得英特尔可以持续高投入到CISC处理器的研发,保证了其处理器性能比RISC处理器性能的持续超越,最终赢得了在PC电脑处理器上面的霸主地位。
在八十年代,PC处理器市场上还有摩托罗拉、IBM、SUN、SGI、DEC和HP都在生产自己的RISC处理器,但由于彼此竞争,大打价格战,最终无力与英特尔的CISC处理器竞争,纷纷倒戈。幸好,RISC的微光仍在ARM保存。
(1990年,ARM公司初创时期的办公地点)
1990年新成立的ARM公司,只有12名工程师参与,其中还并不包括开山鼻祖的威尔逊和福巴尔。办公地点是只是一座位于剑桥的谷仓。Acorn公司和苹果公司各占43%的股份,VLSI占了剩余股份,并成为ARM的半导体代工者,也成为第一个获得ARM授权的芯片厂商。
(搭载ARM-6处理器的苹果Newton 掌上电脑)
1993年,ARM在和苹果合作开发一款搭载ARM处理器的苹果Newton MessagePad(一款过于超前的产品)中再次遭遇市场的冷遇。ARM公司意识到企业的成功不能依赖个别产品。
当时,ARM从摩托罗拉青睐的新任CEO鲁宾·沙克斯比创造性地推出了IP授权的商业模式。此后,芯片生产厂商只需从ARM公司获得ARM处理器的授权,并支付前期许可费和后期生产芯片的专利使用费,就可以获得RISC指令集处理器的IP版权以及绝大多数的收益。而ARM公司也不用在承受产品开发失败或销售不利而带来的经营风险。
这是一个互惠、双赢的商业创新,也为以后RISC芯片产业高度分工和加速发展奠定了基础。
1993年,ARM与德州仪器、三星、夏普等半导体巨头的合作,为ARM的推广树立了声誉,也证实了IP授权的商业模式的可行性。
(摩托罗拉6110搭载了ARM-7处理器)
此时,ARM的发展也赶上了移动设备革命的“天时地利”。当时诺基亚6110成为第一部采用ARM处理器的GSM手机。为符合诺基亚减少内存的要求,ARM专门开发了16位的定制指令集,大幅缩减了内存,并最终由德州仪器生产和出售给了诺基亚。而6110上市获得了极大的成功。高通、飞思卡尔、DEC 相继加入ARM-7的授权阵营,后来授权给超过170家公司,至今一共生产出超过100亿颗芯片。ARM-7也成为ARM在移动处理器的旗舰系列。
1998年,ARM公司同时在美国和英国上市。短短几个月内,ARM就成为估值超过10亿美元的上市公司,放到今天就是妥妥的独角兽,但这仅仅只是ARM腾飞的开始。
乘势生长:ARM开启了移动时代
尽管在2001年,互联网行业市场的股市崩盘,整个行业股价大幅下跌,ARM收入也出现锐减,但是ARM并未元气大伤。这一年,沃伦·伊斯特接替萨克斯比,成为ARM控股的CEO,ARM也开始实施未来五年的路线图计划。ARM成为RISC处理器架构标准的目标正在实现。
(ARM Cortex-A15芯片)
上市之后的ARM,有了充足的资金开启更高端的处理器的设计。2001年,ARM-926EJ-S推出,这一IP被授权给超过100家公司,出货量达50亿颗。此后,ARM又陆续推出了ARM-10、ARM-11系列架构。2005年,意识到除了像ARM-9这样的高端处理器市场,还有像低成本低功耗的微控制器市场。所以,ARM又提出了Cortex产品线,并将处理器架构分为:
面向高性能的 Cortex-A、面向实时控制场景的 Cortex-R,以及面向微控制器市场的 Cortex-M。
此后,随着iPhone智能手机的热销以及Android系统的推出,全球进入了智能移动手机的时代,而ARM也即将随着苹果以及一众安卓厂商和移动芯片厂商的支持,而登上移动终端时代的“王座”。由于ARM处理器占据移动终端设备市场90%的份额,ARM每年的出货量从2013年的每年100亿台上涨到现在每年的200亿台。
(ARM的合作生态)
至此,ARM终于走出了一条与英特尔完全不同商业模式、不同市场定位的道路。二者似乎本应在各自的市场领域发展,可以“井水不犯河水”。但随着PC电脑市场的饱和与移动互联时代到来,英特尔早已把注意力投向移动设备领域,而ARM也将自己的处理器伸向英特尔占据的服务器市场。
英特尔早在1997年就从DEC手中买下了StrongARM授权,并把StrongARM升级为Xscale。然而英特尔并没有为XScale采用高度集成的设计模式,导致性能强悍但是功耗过高。2005年,英特尔又拒绝了来自苹果的处理器设计订单,次年因为业务收缩,将XScale出售,等于将占领移动处理器的机会拱手相让,错过了马上到来的移动时代。
等到英特尔再次意识到ARM的威胁之后,想以x86设计的嵌入式Soc——Atom来迎战ARM生态的时候,只能是一败涂地了。
再次开战的这一年是2011年,当时微软宣布,在下一版的视窗操作系统正式支持ARM处理器,这让英特尔x86处理器的市场地位开始发生动摇。到2016年,由于巨额亏损,英特尔停掉了Atom生产线,而ARM芯片的历史出货量达到了惊人的1000亿。当然,ARM生态在向x86系统所占据的服务器市场的进攻中,也是屡屡折戟沉沙。
(ARM面向高性能服务器应用的Neoverse芯片)
现在,在移动终端市场,体量巨大的英特尔也难以对ARM产生任何威胁。随着移动设备和PC设备的边界日益模糊,数据中心也在寻求更多样、更高性能、低耗能的服务器出现,ARM却再一次向PC处理器芯片、服务器芯片以及AI芯片发起新的冲锋。
同时,随着万物互联的智能时代的到来,越来越多的物联网设备都将搭载不止一颗的嵌入式芯片。而这也正是ARM所擅长领域,不过这也是英特尔想要在未来重点突围的新战场。
此后,我们将看到在ARM生态高歌猛进当中,将在英特尔所坚守的领域以及正在到来的智能物联的领域,展开一场全面战争。
ARM成功的“天时地利人和”
ARM之所以能够取得今天的市场地位和商业成就,主要是归功于其建立的创新商业模式——IP授权模式。
与英特尔的重资产、高利润的垄断模式不同,在移动终端领域中,低成本、微利润的处理器是无法支撑这一重度投入模式的。而ARM开创的授权模式,将芯片的架构、设计(Fabless,如高通、英伟达、华为海思)和生产(Foudary,如台积电)分开,既分摊了成本,又提高了生产效率和新工艺迭代的速度,从而也形成了日益繁荣的ARM生态。
(采用ARM架构的苹果芯片)
ARM公司根据芯片设计公司不同的需求和能力,提供了三种不同的对外授权模式,包括:
第一种,处理器授权。买下IP的芯片设计企业只需按照ARM设计好的芯片图纸生产即可。
第二种,处理器优化包和物理IP包授权。芯片设计企业可以直接拿到一系列设计方案,完成芯片的生产,但是自由度更低,处理器类型、代工厂和工艺都由ARM规定好了。
第三种,架构和指令集授权。针对实力雄厚的芯片设计企业,如苹果、高通、三星和华为海思,他们可以直接购买ARM的架构和指令集,自行设计和ARM指令集兼容的处理器。
目前,ARM在全球拥有1000多家处理器授权合作企业、320家处理器优化包和物理IP包授权伙伴,15家架构和指令集授权企业。在众多授权企业的支持下,ARM处理器2015年的出货量达到了150亿个,历史出货总量超过了1000亿个。
ARM每次在研发新一代处理器IP时,最多会挑选三家合作伙伴。这些被选中的公司能更早地了解ARM的设计,会在新产品研发上占据领先地位,但它们也要帮助ARM进行调试、测试,并向ARM提供反馈,ARM也因此能够确保顺利研发,加快应用的速度。
而在盈利模式上,ARM的利润完全依赖IP授权,利润完全取决于授权人、伙伴、客户能卖出的芯片数量,这样就与芯片的设计、生产、销售的企业紧密绑定,合力实现产品的利益最大化,实现共赢。
在处理器发展方向的选择上,ARM所选择的RISC指令集处理器曾在与英特尔选择的CISC指令集交锋中暂时落败,但无疑RISC代表的低功耗的趋势更符合移动时代的发展方向。
现在,RISC指令集的这一优势更加明显。也就是处理器在低功耗的情况下努力突破性能的极限,要比在保持性能提升的情况下降低能耗,是一件更有优势的进程。ARM的处理器芯片向英特尔的处理器芯片领域扩展,正是这一趋势的最好注脚。
总结以上,可以看到,移动时代爆发代表的天时,RISC先进性能代表的地利、ARM创新商业模式代表的人和,共同构成了今天ARM的繁荣生态和全球事业版图。
本内容为作者独立观点,不代表虎嗅立场。未经允许不得转载,授权事宜请联系[email protected]如对本稿件有异议或投诉,请联系[email protected]
End