The man who foresaw rise of chips with everything
Simon London
In April 1965, Electronics magazine published a four page article by Gordon Moore, then director of research and development at Fairchild Semiconductor. The engineer reported that his lab had created an electronic circuit by etching 50 transistors and resistors on to the surface of a silicon chip.
By 1975, he said, it would be possible to cram 65,000 such components on a chip, creating “integrated” circuits of great complexity.
Six years after the invention of the integrated circuit, this seemed like science fiction. But the article's central prediction that the number of components on a chip would double every year or so turned out to be accurate. While the rate of improvement is these days closer to 18 months, Moore's Law continues to hold.
Today's state-of-the-art integrated circuits comprise 1.7bn transistors.
The implications were profound not only for Mr Moore who went on to co-found Intel and make billions from the technology he pioneered but also for society.
Personal computers, mobile phones, genomic research and computer-aided design would all have been impossible if circuits had to be made the old-fashioned way, by wiring together transistors, resistors and capacitors on circuit boards.
These and other innovations are credited with helping to increase labour productivity and, therefore, boosting economic growth. Economists and policymakers are as likely as technologists to debate the future of Moore's Law.
As he approaches the 40th anniversary of his famous article, Mr Moore is disarmingly unaffected by his enormous wealth and his status as a technology seer.
“I wasn't expecting it to be especially accurate,” confesses the 76-year-old billionaire, who now splits his time between California's Silicon Valley and Hawaii.
He admits to having been deeply embarrassed in the early 70s when his friend Carver Mead, a professor at the California Institute of Technology, coined the term Moore's Law: “I couldn't bring myself to utter it for 20 years.”
Craig Barrett, Intel's current chief executive, is among those who believe the trajectory mapped out by Mr Moore will continue for many years to come.
But skeptics point out that transistors the on/off switches that are the building blocks of integrated circuits are now so small they no longer work as they should.
With each component measuring less than one micron 1/100th the width of a human hair electrical current passes through transistors even when they are off.
This not only causes chips to consume excessive power and generate unwanted heat, it also starts to undermine the basic binary nature of computer circuits.
Mr Moore does not pretend to have any special insight into what the next 40 years will bring: “I've never been able to see more than two or three generations [of chip] ahead without seeing something that looked like a fundamental roadblock.”
But if Moore's Law did run out of road, what would be the economic impact?
Erik Brynjolfsson, professor of economics at MIT's Sloan School of Management, points out that it takes years for companies to take advantage of advances in information technology.
“If Moore's Law ground to a halt tomorrow we would still have 5 to 10 years [of productivity improvements] taking advantage of what we have already got,” he says.
Carver Mead, the Gordon and Betty Moore Professor Emeritus at Caltech, believes that the computing power delivered by Moore's Law has discouraged innovation in other areas of computing, such as software and systems architecture.
He says: “We haven't begun to use the potential of this technology. If we never scaled another device [reduced the size of another transistor] we've got 50 years of innovation based on what we have already got.”
Engineers in other sectors of the IT industry are pursuing exponential improvements of their own. For example, the amount of information that can be stored on magnetic disk drives used in everything from iPods to corporate data warehouses has been increasing at a rate that make Moore's Law look lackadaisical.
Then there is Metcalfe's Law, coined by Robert Metcalfe, founder of networking equipment company 3Com, which says that the usefulness of a network equals the square of the number of users.
On this view, huge benefits will come from increasing the number of people with access to the internet and other communications technology. Raw computing power is not everything.
But this is not to deny Mr Moore's prescience or the remarkable feats of engineering that have enabled chipmakers to remain on the path he set 40 years ago. As the man says himself: “Moore's Law is a violation of Murphy's Law.”
摩尔定律还有下一个40年吗
1965年4月，《电子学》杂志(Electronics)刊登了高登?摩尔(Gordon Moore)撰写的一篇4页文章。高登?摩尔当时是飞兆半导体公司(Fairchild Semiconductor)研发部门的主管。这位工程师报告说，他的实验室通过将50只晶体管和电阻器蚀刻在一张芯片表面，制成一个电子线路。
摩尔说，到1975年，就可能将6.5万只这样的零件密植在一张芯片上，制成高度复杂的集成电路。
当时集成电路问世才6年，摩尔的预测听起来像是科幻小说。但那篇文章的核心预测(即每个芯片可集成的零件差不多每年可增加一倍)被证明是正确的。尽管当今这一技术进步的周期更接近18个月，但“摩尔定律”(Moore’s Law)依然有效。
目前最先进的集成电路含有17亿个晶体管。
摩尔的预言不仅对他本人，而且对整个社会而言都是意义深远的。后来摩尔与他人共同成立了英特尔公司，并通过他所开创的技术创造了数十亿美元的财富。
如果按照旧有方式制造电路，即将晶体管、电阻器和电容器安装在电路板上，那么个人电脑、移动电话、基因研究和计算机辅助设计都不可能问世。
劳动生产率的提高及其所推动的经济增长，都要归功于这类技术创新。如同技术专家一样，经济学家和决策者们也都在探讨摩尔定律的未来。
在那篇著名文章发表40周年前夕，摩尔先生却对巨额财富和“技术先知”称号处之泰然。
“当时我并未指望我的预言会特别准确，”这位76岁的亿万富翁承认。他在加利福尼亚的硅谷和夏威夷两地居住。
他承认，当他的朋友，加州理工学院教授卡弗?米德(Carver Mead )在70年代早期提出“摩尔定律”这一说法时，他深感难为情：“在长达20年的时间中，我无法说出这个名称。”
英特尔公司现任首席执行官克里格?贝瑞特(Craig Barrett)也是摩尔定律的信奉者之一，确信该定律在未来很多年仍将成立。
但怀疑人士指出，晶体管(作为集成电路基本构件的开关)尺寸已经很小，再也无法正常工作。
今天，集成电路各零件的尺寸已经不足1微米，即不到人类头发直径的1/100，以至于在晶体管关闭时，电流仍能通过。
这不仅会导致芯片耗电过多、产生多余热量，而且已经开始危及计算机电路的二进制基本运算机制。
摩尔并未假装自己对未来40年的发展趋势有任何独特见解：“我每一次预见未来2或3代后的芯片，都会看到像是根本性的障碍。”
然而，如果摩尔定律确实寿终正寝，将会有什么经济影响？
麻省理工学院斯隆(Sloan)管理学院教授埃里克?布莱恩约福森(Erik Brynjolfsson)指出，企业利用信息技术取得进步成果需要数年时间。
“假设摩尔定律明天就失效，我们仍可利用现有技术成果在5到10年时间内提高劳动生产率。”他说。
加州理工学院高登和贝蒂(Gordon Betty Moore)基金荣誉教授卡弗?米德认为，按摩尔定律产生的计算能力，阻碍了软件和系统构造等其他领域的创新。
他说：“我们尚未开始利用这种技术的潜力。就算我们不再减小晶体管的体积，我们在现有技术基础上还有50年创新空间。”
目前，其他IT领域的工程师正在以指数级速度推进技术创新。比如，存储技术(应用于从iPod到企业数据库等各类产品中)的发展速度，已经使摩尔定律相形见绌。
此外，网络设备制造商3Com公司的创始人罗伯特?麦特卡夫也提出了“麦特卡夫定律” (Metcalfe’s Law)。该定律指出，网络的有用性(价值)等于网络用户数的平方数。
根据这一观点，增加网络和其他通讯技术的接入用户数量，将获得巨大效益。原始的计算能力并非一切。
然而，这并非否定摩尔的先见之明，也并非否定芯片设计领域取得的巨大成就。40年来，芯片制造商按摩尔预见的发展路径取得技术进步。正如摩尔本人所说：“摩尔定律违反了墨菲法则(Murphy’s Law)。”(编者按：“墨菲法则”认为可能出错的事终将出错。)
