Each year end-users acquire hundreds of millions of various electronic gadgets that are worth hundreds of billions of dollars. The absolute majority of companies have broad product lines with models aiming different types of customers with diverse requirements and income. But it does not take hundreds of devices to change the future of the market forever: the future can be changed with one innovative device, one right or wrong decision or just the progress itself.
While there are market-changing events or products, they do not emerge overnight, they are not traditional milestones, but they are results of evolutionary chains of events that transform into revolutions.
In this article we are going to take a look not only on the products that eventually became strategic inflection points for the industry and catalyzed creation of new types of companies and emergence of new rivals, but also try to understand why did those products materialized at all and which changes did they bring.
2000 - Athlon and Duron Remake AMD
For many years Advanced Micro Devices was a second-source of x86 microprocessors for makers of personal computers and was basically the shadow of its larger rival Intel Corp. For many years the company made clones of Intel-designed processors and hardly could imagine itself as a rival that would offer high-performance leading-edge products. But the last decade changed everything: even though AMD is still a lot smaller than Intel, its microprocessors now compete head-to-head against Intel's chips.
Founded in 1969 by seven ex-Fairchild employees headed by Jerry Sanders, AMD made random access memory (RAM), simplistic logic chips and eventually clones of Intel microprocessors after in 1982 International Business Machines demanded to ensure second source for x86 processors developed by Intel and the chip designer signed an agreement that allowed AMD to produce clones. But in 1986 the larger rival refused to provide technical details about the 80386 microprocessor and after numerous disputes in court, the company had to reverse-engineer both 80386 and 80486-series of microprocessors to create its own. Quite unsurprisingly, as design cycles became shorter, reverse-engineering became inefficient and it was clear that the company needed to design its own chips. The first two generations of AMD's own microprocessors - K5 and K6 - were not exactly competitive from floating point performance point of view, yet, they still were inexpensive alternative to Intel's thanks to the fact that they could be installed into the same sockets.
The seventh-generation AMD microprocessor should not only offer performance on par or higher compared to Intel's Pentium, but also be fast enough to make mainboard manufacturers produce motherboards only compatible with AMD chips. Knowing that, Jerry Sanders, the co-founder and the chief executive officer of AMD, gradually hired engineers from outside the company since the mid nineties. Following his "people first, profit and products will follow motto", Mr. Sanders hired Dirk Meyer and a team of talented engineers who once co-developed Alpha-series processors at DEC in 1996. That team created the processor announced on June 23, 1999, the AMD Athlon.
AMD started to ship its Athlon processor in August, 1999, and the world quickly discovered that the chip could outperform Intel's Pentium III at the same clock-speed and had better frequency potential. Moreover, Intel's 820 core-logic compatible with the latest Pentium III "Coppermine" processors used expensive RDRAM memory developed by Rambus, which did not provide any tangible performance benefits. A version of the core-logic with memory translator hub (MTH), which allowed to use traditional PC100 SDRAM, appeared to be extremely buggy. The result was clear: AMD systems were faster, less expensive and more reliable.
Unfortunately for AMD, the new central processing unit came in slot form-factor, which meant that mainboards for Athlon were expensive to make. In addition, AMD's 750 core-logic was not absolutely the best chipset ever designed and Via Technologies' KX133 was nowhere to be found. To make the matters worse, manufacturers of motherboards were reluctant to manufacture platforms for Athlon (since they did not want to ruin the relationship with Intel, an explanation I will hear in 2005 from a server maker, who back then did not make Opteron-based servers), which greatly slowed down adoption of the chip by the masses. But the main task was achieved: AMD's CPUs could outperform Intel's without any technical problems.
The Sunnyvale, California-based AMD understood the issues with the first-generation Athlon very well and made its homework with the release of the second-generation Athlon code-named Thunderbird in mid-2000.
- Firstly, the new chips came in traditional socket form-factor;
- Secondly, they featured integrated L2 cache, something that ensured linear clock-speed scalability of processor's performance;
- Thirdly, by the time the new microprocessors hit the market AMD ensured that Via's next-generation KT133 chipset was available widely;
- Fourthly, AMD made sure - and great performance of the original Athlon was a help here - that different mainboards for the new CPUs from all important manufacturers were available widely.
But AMD did not only launch a better AMD Athlon, it also released the AMD Duron, a low-cost chip without L2 cache at all, that did not only blew away Celeron, but actually managed to rival much more expensive Pentium III.
The launch of the AMD socket A platform was a huge win. End-users greatly gained confidence in AMD and its products. A number of leading computer makers adopted both Athlon and Duron. Seeing the potential of AMD, software designers started to tune their applications for the company's chips. Mainboard manufacturers initiated production of high-performance motherboards for AMD microprocessors, a never before seen scenario. By late 2000 AMD Athlon processors have received over 70 awards from journalists and market analysts.
The material results were terrific. In early January, 2001, AMD reported annual sales for 2000 at $4.644 billion, a 63% increase year-over-year, as well as net income of $983.026 million, another record for the company.
"Our technology and manufacturing organizations distinguished themselves from the competition by executing nearly flawlessly. [...] Despite the PC slowdown, AMD gained market share in the PC processor arena, with more than 26.5 million total units sold in 2000. We believe we gained three points of market share in units during the year, to approximately 17% of the worldwide market for PC processors," said W.J. Sanders III, chairman and chief executive officer of AMD back then.
"Athlon and Duron processors in the socket A version were just loved by enthusiasts. Probably, they, along with Intel's Celeron at the time may be called the beginning of the era of mass overclocking - and partly due to this fact AMD owes its success. Socket A processors were very friendly to overclocking both because they were easy to overclock and because they were really overclockable since AMD artificially limited their frequency. Many remember the famous 'golden bridges' on the surface of the processors' substrate, modification of which could easily and naturally change the multiplier of these processors," said Ilya Gavrichenkov, CPU analyst at X-bit labs.
It will take Intel to introduce a new processor micro-architecture called Netburst and wait for over a year for it to see its potential, drop exclusive support for RDRAM and reconsider a number of things to return the performance crown in mid-2002. AMD will have to face numerous new challenges on the market and redevelop itself in the following years completely for several times. There will be wins, losses, exciting introductions that will also bring major changes.
But the year 2000 is the year of the triumph of AMD and the market of desktop microprocessors: no more Intel's top chips for the price of over $1000 per unit; no more "Intel only" approach of computer enthusiasts; no more claims about low performance of AMD processors by anyone. The world now has two suppliers of microprocessors that are equally good.