Gene Myron Amdahl

In a long and fruitful career as a computer designer Gene Myron Amdahl created many innovations and refinements in the design of mainframe computers, the hefty workhorses of the data processing industry from the 1950s through the 1970s. Amdahl was born on November 16, 1922, in Flandreau, South Dakota. Amdahl did his college work in electrical engineering and physics. When his studies were interrupted by World War II, he served as a physics instructor for an army special training program and then joined the navy, where he taught electronics until 1946. He then returned to school, receiving his B.S. degree from South Dakota State University in 1948 and his doctorate in physics at the University of Wisconsin in 1952. As a graduate student, Amdahl worked on a problem involving the forces binding together parts of a simple atomic nucleus. He and two fellow students spent a month performing the necessary computations with calculators and slide rules. Amdahl realized that if physicists were going to be able to move on to more complex problems they would need greater computing resources. He therefore designed a computer called the WISC (Wisconsin Integrally Synchronized Computer). This computer used a sophisticated procedure to break calculations into parts that could be carried out on separate processors, making it one of the earliest examples of the parallel computing techniques found in today’s computer processors.
In 1952, Amdahl went to work for IBM, which was beginning the effort that would lead to its dominating the business computer industry by the end of the decade. Amdahl worked with the team that designed the IBM 704. The 704 improved upon the 701, the company’s first successful mainframe, by adding many new internal programming instructions, including the ability to perform floating point calculations (involving numbers that have decimal points). The machine also included a fast, high-capacity magnetic core memory that let the machine retrieve data more quickly during calculations. In November 1953, Amdahl became the chief project engineer for the 704.
On the heels of that accomplishment, new opportunities seemed to be just around the corner. Although IBM had made its reputation in business machines, it was also interested in the market for specialized computers for scientists. Amdahl helped design the IBM 709, an extension of the 704 designed for scientific applications. When IBM proposed extending the technology by building a powerful new scientific computer called STRETCH, Amdahl eagerly applied to head the new project. However he ended up on the losing side of a corporate power struggle, and did not receive the post. He left IBM at the end of 1955.
Amdahl then worked for several small data processing companies. He helped design the RW440, a minicomputer used for industrial process control. This period gave Amdahl some experience in dealing with the problems of startup businesses, experience he would call upon later when he started his own company. In 1960, Amdahl rejoined IBM and soon was involved in several design projects. The one with the most lasting importance was the IBM System/360, which would become the most ubiquitous and successful mainframe computer of all time. In this project, Amdahl further refined his ideas about making a computer’s central processing unit more efficient. He designed logic circuits that enabled the processor to analyze the instructions waiting to be executed (the “pipeline”) and determine which instructions could be executed immediately and which would have to wait for the results of other instructions. He also used a cache, or special memory area in which the instructions that would be needed next could be stored ahead of time so they could be retrieved quickly from high-speed storage. Today’s desktop personal computers (PCs) use these same ideas to get the most out of their chips’ capabilities. The problem of parallel computing is partly a problem of designing appropriate hardware and partly a problem of writing (or rewriting) software so its instructions can be executed simultaneously. It is often difficult to predict how much a parallel computing arrangement will improve upon using a single processor and conventional software. Amdahl created a formula called Amdahl’s law, which attempts to answer that question. In simple terms, Amdahl’s law says that the advantage gained from using more processors gradually declines as more processors are added. The amount of improvement is also proportional to how much of the calculation can be broken down into parts that can be run in parallel. As a result, some kinds of programs can run much faster with several processors being used simultaneously, while other programs may show little improvement.
As a designer, Amdahl coupled hard work with the ability to respond to sudden bursts of intuition. “Sometimes,” he recalled to author Robert Slater, “I wake up in the middle of the night and I’ll be going 60 miles an hour on the way to a solution. I see a mental picture of what is going on and I dynamically operate that in my mind.” In 1965, Amdahl was awarded a five-year IBM fellowship that allowed him to study whatever problems interested him. He also helped establish IBM’s Advanced Computing Systems Laboratory in Menlo Park, California, which he directed. However, Amdahl became increasingly frustrated with what he thought was IBM’s toorigid approach to designing and marketing computers. IBM insisted on basing the price of a new computer not on how much it cost to produce, but on how fast it could calculate.
Amdahl wanted to build much more powerful computers—what would soon be called “supercomputers.” But if IBM’s policy were followed, these machines would be so expensive that virtually no one would be able to afford them. Thus at a time when increasing miniaturization was leading to the possibility of much more powerful machines, IBM did not seem to be interested in building them. The computer giant seemed to be content to gradually build upon its financially successful 360 line (which would become the IBM 370 in the 1970s). Amdahl therefore left IBM in 1970, later recalling to Slater that he left IBM that second time “because I wanted to work in large computers. . . . I’d have had to change my career if I stayed at IBM—for I wanted personal satisfaction.”
To that end, in 1970 he founded the Amdahl Corporation. Amdahl resolved to make computers that were more powerful than IBM’s machines, but would be “plug compatible” with them, allowing them to use existing hardware and software. Business users who had already invested heavily in IBM equipment could thus buy Amdahl’s machines without fear of incompatibility. Since IBM was known as “Big Blue,” Amdahl decided to become “Big Red,” painting his machines accordingly.
Amdahl would later recall his great satisfaction in “getting those first computers built and really making a difference, seeing it completely shattering the control of the market that IBM had, causing pricing to come back to realistic levels.” Amdahl’s critics sometimes accused him of having unfairly used the techniques and knowledge that he had developed at IBM, but he has responded by pointing to his later technical innovations. In particular, he was able to take advantage of the early developments in integrated electronics to put more circuits on a chip without making the chips too small, and thus too crowded for placing the transistors. After it was introduced in 1975, the Amdahl 470 series of machines, doubled in sales in each of its first three years. Thanks to the use of largerscale circuit integration, Amdahl could sell machines with superior technology to that of the IBM 360 or even the new IBM 370, and at a lower price. IBM responded belatedly to the competition, making more compact and faster processors, but Amdahl met each new IBM product with a faster, cheaper alternative. However, IBM also countered by using a sales technique that opponents called FUD—fear, uncertainty, and doubt. IBM salespersons promised customers that IBM would soon be coming out with much more powerful and economical alternatives to Amdahl’s machines. As a result, many potential customers were persuaded to postpone purchasing decisions and stay with IBM. Amdahl Corporation began to falter, and Gene Amdahl gradually sold his stock and left the company in 1980.
Amdahl then tried to repeat his early success by starting a new company called Trilogy. The company promised to build much faster and cheaper computers than those offered by IBM or Amdahl. He believed he could accomplish this by using the new, very-large-scale integrated silicon wafer technology, in which circuits were deposited in layers on a single chip rather than being distributed on separate chips on a printed circuit board. However, the problem of dealing with the electrical characteristics of such dense circuitry, as well as some design errors, somewhat crippled the new computer design. Amdahl also found that the aluminum substrate that connected the wafers on the circuit board was causing short circuits. Even weather, in the form of a torrential rainstorm, conspired to add to Amdahl’s problems by flooding a chip-building plant and contaminating millions of dollars’ worth of chips. Amdahl was forced to repeatedly delay the introduction of the new machine, from 1984 to 1985 to 1987. He attempted to infuse new technology into his company by buying Elxsi, a minicomputer company, but Trilogy never recovered.
After the failure of Trilogy, Amdahl undertook new ventures in the late 1980s and 1990s, including Andor International, an unsuccessful developer of minicomputers, and Commercial Data Servers (CDS), which is trying to compete with IBM in the low-priced end of the mainframe market.
Amdahl has received many industry awards, including “Data Processing Man of the Year” from the Data Processing Management Association (1976) and the Harry Goode Memorial Award from the American Federation of Information Processing Societies.