AMD 的兴起、衰落和复兴

注：机翻，未校。

---

The Rise, Fall and Revival of AMD

By Nick Evanson and Graham Singer June 29, 2020

AMD is one of the oldest designers of large scale microprocessors and has been the subject of polarizing debate among technology enthusiasts for nearly 50 years. Its story makes for a thrilling tale – filled with heroic successes, foolhardy errors, and a close shave with rack and ruin. Where other semiconductor firms have come and gone, AMD has weathered many storms and fought numerous battles, in boardrooms, courts, and stores.
AMD 是最古老的大规模微处理器设计商之一，近 50 年来一直是技术爱好者争论不休的主题。它的故事构成了一个惊心动魄的故事 —— 充满了英勇的成功、鲁莽的错误，以及与机架和废墟的近距离剃须。其他半导体公司来来去去，而 AMD 在会议室、法院和商店中经受住了许多风暴，打了无数场仗。

In this feature we’ll revisit the company’s past, examine the twists and turns in the path to the present, and wonder at what lies ahead for this Silicon Valley veteran.
在本专题中，我们将重温公司的过去，审视通往现在的曲折之路，并思考这位硅谷资深人士的未来。

Editor’s note: The third edition of this article was published at the end of 2022, see the most up-to-date version here.
编者按：本文的第三版于 2022 年底发布，请在此处查看最新版本。

The rise to fame and fortune 名利双收

To begin our story, we need to roll back the years and head for America and the late 1950s. Thriving after the hard years of World War II, this was the time and place to be if you wanted experience the forefront of technological innovation.
要开始我们的故事，我们需要倒退几年，前往美国和 1950 年代后期。在第二次世界大战的艰苦岁月之后蓬勃发展，如果您想体验技术创新的前沿，那么这里就是您的好去处。

Companies such as Bell Laboratories, Texas Instruments, and Fairchild Semiconductor employed the very best engineers, and churned out numerous firsts: the bipolar junction transistor, the integrated circuit, and the MOSFET (metal oxide semiconductor field effect transistor).
贝尔实验室、德州仪器和 Fairchild Semiconductor 等公司聘请了最优秀的工程师，并创造了许多第一：双极结型晶体管、集成电路和 MOSFET（金属氧化物半导体场效应晶体管）。

Fairchild engineers, circa 1960 – Gordon Moore is far left, Robert Noyce is in the middle foreground
Fairchild engineers ，约 1960 年 – Gordon Moore 在最左边，Robert Noyce 在中间前景

These young technicians wanted to research and develop ever more exciting products, but with cautious senior managers mindful of the times when the world was fearful and unstable, frustration amongst the engineers build a desire to strike out alone.
这些年轻的技术人员希望研究和开发更多令人兴奋的产品，但随着谨慎的高级管理人员意识到世界充满恐惧和不稳定的时代，工程师们的挫败感产生了独自出击的愿望。

And so, in 1968, two employees of Fairchild Semiconductor, Robert Noyce and Gordon Moore, left the company and forged their own path. N M Electronics opened its doors in that summer, to be renamed just weeks later as Integrated Electronics – Intel, for short.
因此，在 1968 年，Fairchild Semiconductor 的两名员工 Robert Noyce 和 Gordon Moore 离开了公司，开辟了自己的道路。NM Electronics 在那年夏天开业，几周后更名为 Integrated Electronics—— 简称 Intel。

Others followed suit and less than a year later, another 8 people left and together they set up their own electronics design and manufacturing company: Advanced Micro Devices (AMD, naturally).
其他人纷纷效仿，不到一年后，又有 8 人离开，他们一起成立了自己的电子设计和制造公司：Advanced Micro Devices（自然是 AMD）。

The group was headed by Jerry Sanders, Fairchild’s former director of marketing, They began by redesigning parts from Fairchild and National Semiconductor rather than trying to compete directly with the likes of Intel, Motorola, and IBM (who spent significant sums of money on research and development of new integrated circuits).
该小组由 Fairchild 的前营销总监 Jerry Sanders 领导，他们首先重新设计了 Fairchild 和 National Semiconductor 的零件，而不是试图与英特尔、摩托罗拉和 IBM 等公司直接竞争（他们花费了大量资金研发新的集成电路）。

From these humble beginnings, and headquartered in Silicon Valley, AMD offered products that boasted increased efficiency, stress tolerances, and speed within a few months. These microchips were designed to comply with US military quality standards, which proved a considerable advantage in the still-young computer industry, where reliability and production consistency varied greatly.
从这些不起眼的开始，总部位于硅谷的 AMD 在几个月内提供的效率、抗压能力和速度都得到提升的产品。这些微芯片的设计符合美国军用质量标准，这在仍然年轻的计算机行业中被证明是一个相当大的优势，因为该行业的可靠性和生产一致性差异很大。

AMD’s first copycat CPU – the Am9080. Image: Wikipedia
AMD 的第一款山寨 CPU – Am9080。图片： 维基百科上的数据

By the time Intel released their first 8-bit microprocessor (the 8008) in 1974, AMD was a public company with a portfolio of over 200 products – a quarter of which were their own designs, including RAM chips, logic counters, and bit shifters. The following year saw a raft of new models: their own Am2900 integrated circuit (IC) family and the 2 MHz 8-bit Am9080, a reverse-engineered copy of Intel’s successor to the 8008. The former was a collection of components that are now fully integrated in CPUs and GPUs, but 35 years ago, arithmetic logic units and memory controllers were all separate chips.
到 Intel 于 1974 年发布他们的第一个 8 位微处理器 （8008） 时，AMD 已经是一家上市公司，拥有 200 多种产品组合，其中四分之一是他们自己的设计，包括 RAM 芯片、逻辑计数器和移位器。次年，他们推出了大量新型号：他们自己的 Am2900 集成电路 （IC） 系列和 2 MHz 8 位 Am9080，这是英特尔 8008 继任者的逆向工程副本。前者是现在完全集成在 CPU 和 GPU 中的组件集合，但在 35 年前，算术逻辑单元和内存控制器都是独立的芯片。

The blatant plagiarism of Intel’s design might seem to be somewhat shocking by today’s standards, but it was par for the course in the fledgling days of microchips.
以今天的标准来看，公然剽窃英特尔的设计似乎有些令人震惊，但在微芯片刚刚起步的时代，这已经是正常的了。

The blatant plagiarism of Intel’s design might seem to be somewhat shocking by today’s standards, but it was par for the course in the fledgling days of microchips. The CPU clone was eventually renamed as the 8080A, after AMD and Intel signed a cross-licensing agreement in 1976. You’d imagine this would cost a pretty penny or two, but it was just $325,000 ($1.65 million in today’s dollars).
以今天的标准来看，公然剽窃英特尔的设计似乎有些令人震惊，但在微芯片刚刚起步的时代，这已经是正常的了。在 AMD 和 Intel 于 8080 年签署交叉许可协议后，CPU 克隆最终更名为 1976A。您会想象这会花费一两美分，但实际上只是 325,000 美元（按今天的美元计算为 165 万美元）。

The deal allowed AMD and Intel to flood the market with ridiculously profitable chips, retailing at just over $350 or twice that for ‘military’ purchases. The 8085 (3 MHz) processor followed in 1977, and was soon joined by the 8086 (8 MHz). In 1979 also saw production begin at AMD’s Austin, Texas facility.
这笔交易使 AMD 和英特尔能够用利润丰厚的芯片充斥市场，零售价略高于 350 美元，是 “军用” 购买的两倍。8085 （3 MHz） 处理器于 1977 年推出，很快又推出了 8086 （8 MHz）。1979 年，AMD 位于德克萨斯州奥斯汀的工厂也开始生产。

When IBM began moving from mainframe systems into so-called personal computers (PCs) in 1982, the outfit decided to outsource parts rather than develop processors in-house. Intel’s 8086, the first ever x86 processor, was chosen with the express stipulation that AMD acted as a secondary source to guarantee a constant supply for IBM’s PC/AT.
当 IBM 于 1982 年开始从大型机系统转向所谓的个人计算机 （PC） 时，该公司决定将零件外包，而不是在内部开发处理器。英特尔的 8086 是有史以来第一个 x86 处理器，被选中时明确规定 AMD 作为次要来源，以保证 IBM 的 PC/AT 的持续供应。

Any color, as long as it’s beige. IBM’s 5150 PC from 1981
任何颜色，只要是米色即可。1981 年的 IBM 5150 PC

A contract between AMD and Intel was signed in February of that year, with the former producing 8086, 8088, 80186, and 80188 processors – not just for IBM, but for the many IBM clones that proliferated (Compaq being just one of them). AMD also started manufacturing the 16-bit Intel 80286, badged as the Am286, towards the end of 1982.
当年 2 月，AMD 和英特尔签署了一份合同，前者生产 8086、8088、80186 和 80188 处理器 —— 不仅适用于 IBM，还用于激增的许多 IBM 克隆产品（康柏只是其中之一）。AMD 还在 16 年底开始制造 80286 位 Intel 1986，标记为 Am282。

This was to become the first truly significant desktop PC processor, and while Intel’s models generally ranged from 6 to 10 MHz, AMD’s started at 8 MHz and went as high as 20 MHz. This undoubtedly marked the start of the battle for CPU dominance between the two Silicon Valley powerhouses; what Intel designed, AMD simply tried to make better.
这将成为第一个真正重要的台式 PC 处理器，虽然 Intel 的型号通常从 6 到 10 MHz 不等，但 AMD 的型号从 8 MHz 开始，一直到 20 MHz。这无疑标志着两家硅谷强国之间争夺 CPU 主导地位的战斗的开始；Intel 设计的东西，AMD 只是想做得更好。

This period represented a huge growth of the fledgling PC market, and noting that AMD had offered the Am286 with a significant speed boost over the 80286, Intel attempted to stop AMD in its tracks. This was done by excluding them from gaining a licence for the next generation 386 processors.
这一时期代表了新兴 PC 市场的巨大增长，并注意到 AMD 为 Am286 提供了比 80286 更快的速度，英特尔试图阻止 AMD 的发展。这是通过排除他们获得下一代 386 处理器的许可证来实现的。

AMD sued, but arbitration took four and a half years to complete, and while the judgment found that Intel was not obligated to transfer every new product to AMD, it was determined that the larger chipmaker had breached an implied covenant of good faith.
AMD 提起诉讼，但仲裁花了四年半时间才完成，虽然判决认定英特尔没有义务将每款新产品都转让给 AMD，但确定这家大型芯片制造商违反了默示的诚信契约。

Intel’s licence denial occurred during a critical period, right as IBM PC’s market was ballooning from 55% to 84%. Left without access to new processor specifications, AMD took over five years to reverse-engineer the 80386 into the Am386. Once completed, it proved once more to be more than a match for Intel’s model. Where the original 386 debuted at just 12 MHz in 1985, and later managed to reach 33 MHz, the top-end version of the Am386DX launched in 1989 at 40 MHz.
英特尔的许可证被拒绝发生在一个关键时期，当时 IBM PC 的市场正从 55% 膨胀到 84%。由于无法获得新的处理器规格，AMD 花了五年多的时间将 80386 逆向工程为 Am386。一旦完成，它再次证明它与 Intel 的模型相得益彰。最初的 386 在 1985 年首次亮相时时频率仅为 12 MHz，后来成功达到 33 MHz，而 Am386DX 的高端版本于 1989 年以 40 MHz 推出。

The Am386’s success was followed by the release of 1993’s highly competitive 40 MHz Am486, which offered roughly 20% more performance than Intel’s 33 MHz i486 for the same price. This was to be replicated throughout the entire 486 line up, and while Intel’s 486DX topped out at 100 MHz, AMD offered (somewhat predictably at this stage) a snappier 120 MHz option. To better illustrate AMD’s good fortune in this period, the company’s revenue doubled from just over $1 billion in 1990 to well over $2 billion in 1994.
Am386 的成功之后，1993 年发布了极具竞争力的 40 MHz Am486，在相同的价格下，它的性能比 Intel 的 33 MHz i486 高出约 20%。这将在整个 486 系列中复制，虽然英特尔的 486DX 最高为 100 MHz，但 AMD 提供了（在这个阶段有点可以预见)更敏捷的 120 MHz 选项。为了更好地说明 AMD 在这一时期的好运气，该公司的收入从 1990 年的略高于 10 亿美元翻了一番，到 1994 年达到了 20 亿美元以上。

In 1995, AMD introduced the Am5x86 processor as a successor to the 486, offering it as a direct upgrade for older computers.
1995 年，AMD 推出了 Am5x86 处理器作为 486 的继任者，将其作为旧计算机的直接升级。

In 1995, AMD introduced the Am5x86 processor as a successor to the 486, offering it as a direct upgrade for older computers. The Am5x86 P75+ boasted a 150 Mhz frequency, with the ‘P75’ referencing performance that was similar to Intel’s Pentium 75. The ‘+’ signified that the AMD chip was slightly faster at integer math than the competition.
1995 年，AMD 推出了 Am5x86 处理器作为 486 的继任者，将其作为旧计算机的直接升级。Am5x86 P75+ 拥有 150 Mhz 的频率，“P75” 参考性能类似于 Intel 的 Pentium 75。“+” 表示 AMD 芯片在整数数学运算方面比竞争对手略快。

To counter this, Intel altered its naming conventions to distance itself from products by its rival and other vendors. The Am5x86 generated significant revenue for AMD, both from new sales and for upgrades from 486 machines. As with the Am286, 386 and 486, AMD continued to extend the market scope of the parts by offering them as embedded solutions.
为了解决这个问题，英特尔改变了其命名约定，以使其与竞争对手和其他供应商的产品保持距离。Am5x86 为 AMD 带来了可观的收入，包括新销售和 486 台机器的升级。与 Am286、386 和 486 一样，AMD 通过提供嵌入式解决方案来继续扩大这些部件的市场范围。

March 1996 saw the introduction of its first processor, developed entirely by AMD’s own engineers: the 5k86, later renamed to the K5. The chip was designed to compete with the Intel Pentium and Cyrix 6x86, and a strong execution of the project was pivotal to AMD – the chip was expected to have a much more powerful floating point unit than Cyrix’s and about equal to the Pentium 100, while the integer performance targeted the Pentium 200.
1996 年 3 月，它推出了第一款完全由 AMD 自己的工程师开发的处理器：5k86，后来更名为 K5。该芯片旨在与 Intel Pentium 和 Cyrix 6x86 竞争，该项目的强大执行对 AMD 至关重要 —— 预计该芯片将具有比 Cyrix 更强大的浮点单元，大约等于 Pentium 100，而整数性能针对 Pentium 200。

A false color image of the K5 die. Image: Wikipedia

Ultimately, it was a missed opportunity, as the project was dogged with design and manufacturing issues. These resulted in the CPU not meeting frequency and performance goals, and it arrived late to market, causing it to suffer poor sales.
最终，这是一个错失的机会，因为该项目受到设计和制造问题的困扰。这导致 CPU 无法达到频率和性能目标，并且上市时间较晚，导致销售不佳。

By this time, AMD had spent $857 million in stock on NexGen, a small fabless chip (design-only) company whose processors were made by IBM. AMD’s K5 and the developmental K6 had scaling issues at higher clock speeds (~150 MHz and above) while NexGen’s Nx686 had already demonstrated a 180 MHz core speed. After the buyout, the Nx686 became AMD’s K6 and the development of the original chip was consigned to the scrapyard.
此时，AMD 已经在 NexGen 上花费了 8.57 亿美元的股票，NexGen 是一家小型无晶圆厂芯片（仅设计）公司，其处理器由 IBM 制造。AMD 的 K5 和开发的 K6 在更高的时钟速度（~150 MHz 及以上）下存在扩展问题，而 NexGen 的 Nx686 已经展示了 180 MHz 的核心速度。买断后，Nx686 成为 AMD 的 K6，原始芯片的开发被委托给废料场。

The K6-2 introduced AMD’s 3DNow! SIMD (single instruction, multiple data) instruction set.
K6-2 引入了 AMD 的 3DNow！SIMD（单指令、多数据)指令集。

AMD’s rise reflected Intel’s decline, from the early beginnings of the K6 architecture, which was pitted against Intel’s Pentium, Pentium II and (largely rebadged) Pentium III. The K6 produced a quickening of AMD’s success, owing its existence and capabilities to an ex-Intel employee, Vinod Dham (a.k.a. the “Father of Pentium”), who left Intel in 1995 to work at NexGen.
AMD 的崛起反映了英特尔的衰落，从 K6 架构的早期开始，它与英特尔的奔腾、奔腾 II 和（大部分重新贴牌的）奔腾 III 相抗衡。K6 加速了 AMD 的成功，这要归功于前英特尔员工 Vinod Dham（又名 “奔腾之父”），他于 1995 年离开英特尔前往 NexGen 工作。

When the K6 hit shelves in 1997, it represented a viable alternative to the Pentium MMX. The K6 went from strength to strength – from a 233 MHz speed in the initial stepping, to 300 MHz for the “Little Foot” revision in January 1998, 350 MHz in the “Chomper” K6-2 of May 1998, and an astonishing 550 MHz in September 1998 with the “Chomper Extended” revision.
当 K6 于 1997 年上架时，它代表了 Pentium MMX 的可行替代品。K6 不断壮大 —— 从最初步进的 233 MHz 速度，到 1998 年 1 月 “Little Foot” 修订版的 300 MHz，1998 年 5 月 “Chomper” K6-2 的 350 MHz，以及 1998 年 9 月 “Chomper Extended” 修订版的惊人 550 MHz。

Image: Wikipedia

The K6-2 introduced AMD’s 3DNow! SIMD (single instruction, multiple data) instruction set. Essentially the same as Intel’s SSE, it offered an easier route accessing the CPU’s floating point capabilities; the downside to this being that programmers needed to incorporate the new instruction in any new code, in addition to patches and compilers needing to be rewritten to utilize the feature.
K6-2 引入了 AMD 的 3DNow！SIMD（单指令、多数据）指令集。与 Intel 的 SSE 基本相同，它提供了一条更简单的途径来访问 CPU 的浮点功能；这样做的缺点是，除了需要重写补丁和编译器才能使用该功能之外，程序员还需要将新指令合并到任何新代码中。

Like the initial K6, the K6-2 represented much better value than the competition, often costing half as much as Intel’s Pentium chips. The final iteration of the K6, the K6-III, was a more complicated CPU, and the transistor count now stood at 21.4 million – up from 8.8 million in the first K6, and 9.4 million for the K6-II.
与最初的 K6 一样，K6-2 比竞争对手的价值要高得多，其成本通常是英特尔奔腾芯片的一半。K6 的最终迭代 K6-III 是一个更复杂的 CPU，晶体管数量现在为 2140 万个，高于第一代 K6 的 880 万个和 K6-II 的 940 万个。

It incorporated AMD’s PowerNow!, which dynamically altered clock speeds according to workload. With clock speeds eventually reaching 570MHz, the K6-III was fairly expensive to produce and had a relatively short life span cut short by the arrival of the K7 which was better sui