Forever second

Anniversary of Intel's main competitor

On May 1, one of the leading processor manufacturers, Advanced Micro Devices (AMD), turns 40 years old. It is difficult to correctly assess its role in the history of the semiconductor industry. One thing is clear: without AMD, the IT world would be different.

Company founder Jeremy Sanders was born on the South Side of Chicago in 1936. In those days, surviving for a teenager from a poor family in this city, which then deservedly bore the title of “stronghold of gangsterism,” was not easy: they say that once, as a result of a conflict with one of the youth gangs, young Jerry was almost killed.

His grandfather helped Jeremy obtain a higher engineering education in electronics (at the University of Illinois). But he did not work as an engineer for long: in 1961 he was lucky enough to get into the marketing department of the famous Fairchild Semiconductor, which in those years occupied approximately the same position in the field of semiconductors as Google does today on the Internet. The company was led by Gordon Moore, whose name is named after the well-known law, and the creator of microchip production technology, Robert Noyce. They would later become the founders of Intel.

Jeremy Sanders' main achievement as head of Fairchild's sales department is that he was one of the first in the world to understand how to sell microchips. In those days, semiconductor chips were very expensive and were mainly used by the military. But if it is impossible to build a house cheaper than the cost of the materials from which it is made, then the cost of a microcircuit tends to zero as production volume increases. This means, Sanders reasoned, the main thing is to ensure a sufficient level of sales.

The canonical story of how he began to sell a transistor costing $100 for a dollar five cents only because it had an almost unlimited potential market: this transistor was ideal for UHF tuners, which became a must-have component just then, has been widely circulated. home TVs.

Fairchild actually managed to capture 90% of the UHF tuner market by the mid-1960s, and in the early 1970s the transistor was already selling for 15 cents. Since then, semiconductors have been traded all over the world this way: both an operational amplifier chip containing several dozen transistors and a microcontroller chip containing several tens of thousands of them have the same price - a dollar.

In 1968, Sherman Farchild, owner of the industrial group that included Fairchild Semiconductor, decided to reform its management, inviting the conservative Lester Hogan, previously vice president of Motorola, as CEO. This marked the beginning of the decline of the company, which had been the center of Silicon Valley for ten years. Its main asset - people - began to melt. Moore and Noyce, as you know, created Intel, in which they were helped by the famous financier Arthur Rock, who took part in the creation of Fairchild (in 1957) and is considered the father of the venture capital model of financing.

But when Jeremy Sanders was going to create his own company called Advanced Micro Devices, Rock refused him, according to legend, with the wording: “All the companies where I lost money were founded by people from marketing departments.” But Robert Noyce invested a certain amount in AMD.

At first everything went as Rock predicted. The first year or two for AMD were, to put it mildly, not very successful. Founded on May 1, 1969 with an authorized capital of $100 thousand, the company did not have a single original engineering idea behind it. Sanders created it in order to sell it the way he liked, and, apparently, he did not have a very clear idea of ​​what exactly to sell. Therefore, a year later, in May 1970, AMD, which then employed 53 employees, still had only 18 items on its product list (licensed copies of other people's designs) and not a single client.

Five years later, sales exceeded 25 million and were growing more than 60% per year - and still nothing original. How did Sanders do it?

He, of course, understood that the ability to “sell” a product to a client alone would not get you far. For a long-term effect, your product must be either better than others, or at least cheaper. This position - "at least cheaper" - is the cornerstone of AMD's policy to this day.

In the mid-1970s, AMD began producing a clone of the i8080 processor, which worked 40% faster than the original (contrary to popular belief, the 8080A was not licensed from Intel, but was developed by reverse engineering - someone else's chip was studied before work on its own began). The legendary agreement between Intel and AMD dates back to 1982, according to which the latter received the rights to produce x86 processors. Intel needed this, among other things, to receive military orders (in the American military-industrial complex there is a rule according to which a supplier cannot be a monopolist, so that in case of emergency the military would not be left without supplies).

But clones of Intel processors from AMD have never been copies. They have always been cheaper, often faster, and often more reliable. In addition, AMD has always been very careful about the quality of production. All products, even for civilian use, were tested according to military standards, and in 1983 the company put forward its own industrial quality standard (INT.STD.1000), which to this day remains the most stringent in some respects.

Another component of AMD's corporate policy is concern for personnel. “First the people, and then the products with profits,” said Sanders. Its employees were among the first in the world to receive a share in the profits.

By the mid-1970s, AMD had its own products. In 1975, it released a series of Am2900 processors that used the so-called. bit-slice topology (bit planes). Using such processors with a given native bit depth (specifically, the Am2901 was four-bit), you could build a system with any required bit depth. These processors, in particular, were used in the PDP-10 computers from DEC, as well as in the famous Xerox Star, which at one time became the prototype for testing almost all the main computer technologies of our day.

At that time, no one was making money on processors. AMD's main product, like Intel's, was memory, both conventional dynamic memory and non-volatile memory (EPROM). This is what later became known as flash memory. Everything changed in the late 1980s, when, firstly, the “Asian tigers” came to the memory market and pushed everyone out, and secondly, when it became finally clear that processors for PCs were a potentially inexhaustible market.

It was then that Intel and AMD quarreled, and a series of mutual lawsuits began, which, with some periods of relative calm and transitions into side storylines, continues to this day. Intel, not without reason, accused AMD of stealing intellectual property, and AMD just as rightly made counterclaims of market monopolization and violation of agreements. Largely because of these disputes, Intel was forced at one time to change the name of the x86 series to Pentium. There is some irony in the fact that in the fall of 1998, Jeremy Sanders was presented with the Robert Noyce Award for Achievement in the Industry, and the award was presented to him by Craig Barrett, who had recently become president of Intel.

The 486 was the last processor that AMD had to fight over in court, despite the fact that most of the code was developed by AMD in-house. And then the companies began to compete for real, developing their own processors for the same market sectors. And although Intel's share of the personal computer market as a whole never fell below two-thirds, AMD managed to overtake its more powerful competitor in a number of positions. For example, at the peak of the “race for megahertz” in 2000, it was the first to overcome the 1 GHz threshold, and in 2003, it released Opteron, the first 64-bit processor for PCs.

The best year for AMD, by all accounts, was 2005. By that time, Jeremy Sanders had already left the corporation. This happened back in 2000, and his successor Hector Rutz, as you can see, adequately held the reins of power for quite a long time. Until he made the biggest strategic miscalculation, not even one, but two at once.

The first concerned the release of multi-core processors, where AMD was decisively late, losing primacy to Intel's Core 2 Duo in 2006. The second turned out to be even more serious: AMD completely missed the boom in mobile devices, in particular netbooks. When the first ASUS Eee PC 701 hit the shelves at the end of 2007 - beginning of 2008, Intel already had the Atom platform almost ready, as if specially designed for such devices. AMD's competing Yukon platform is still not ready. Such a delay, and even in a crisis, leaves AMD no chance to take its rightful place in this sector.

Of course, the company will not disappear anywhere. As a last resort, it will face restructuring, already under the leadership of Dirk Meyer, one of the developers of Athlon processors, who became president of AMD in 2008. For too long, the development of the processor market was determined by the confrontation between Intel and AMD. It will be a shame if this fight ends with a trivial defeat of one of the opponents.





Prepared jointly with iToday.ru

Yuri REVICH, iToday.ru - specially for Vremya Novostei