Issue



The secret of chipmaking success


06/01/2003







Click here to enlarge image

Successfully making high-performance, leading-edge integrated circuits has probably become the highest form of manufacturing art ever. Chipmakers admit that there is a certain amount of black magic when they follow the "copy exact" approach to equipping fabs, since the smallest variation, even spacing between tools or the layout of piping, may mysteriously throw off a carefully orchestrated set of process recipes.

On top of the technical complexity, there is now the problem of diminishing average selling prices (ASPs) for chips. Leading-edge chipmaking is the most challenging manufacturing process in the history of mankind, and even a slight misstep can turn a small profit into a big loss.

That's why we were so amazed recently to find that Samsung was not only able to make a profit on its total business last year — which includes products like DVD players and cell phones as well as ICs — but also actually turned a tidy profit on DRAMs in a period when memory prices were perhaps at their lowest ever. Could it be that a South Korean company — that just a short time ago was known as a maker of second-tier consumer electronics — has now become perhaps the world's most skilled wafer processor?

Because we talk to a lot of companies that deal with fabs all over the world, we already had some idea of the prowess of Samsung. Many years ago, one Korean chipmaker decided to make its fabs as automated as possible, and pushed vendors to integrate their equipment into this automated scheme. But that chipmaker was not willing to pay for the extra engineering effort this entailed. Its vendors knew the efforts would not be very successful. They contrasted this approach to Samsung, where they found a willingness to learn from vendors and to pay for engineering help when needed to achieve better process integration. Later, I met with Samsung engineers when they were in the process of becoming the second source for the Alpha microprocessor once made by DEC. At the time, the Alpha was the most sophisticated microprocessor in the world. Amazingly, they mastered the manufacturing of this complex chip in just six months.

So what is different about the way Samsung goes about qualifying new processes and putting new chip designs into production in its fabs? The answer, as indicated by some of those who have collaborated with Samsung, is careful, meticulous, in-depth engineering studies of all aspects of the processes. Go to them with an instrument that can do a better job of finding, categorizing, and analyzing defects, and they will try it. If it works, they will immediately order a few. Sometimes it takes some science to back up the engineering, particularly when new materials are involved, and Samsung is willing to dig into that as well when necessary.

Since 1996, Yun Jong Yong, an engineer, has led the company. Engineers like Yong know about the value of making careful, detailed studies before plunging ahead. Japanese companies gained considerable fame for manufacturing skills by similar careful, multidisciplinary studies before launching something new. This approach also may allow different groups and departments to recognize potential friction and work out solutions before the problems ever occur. American managers who worked in Japan told me they were ready to climb the walls during this lengthy preparatory phase. Once a project was launched, however, they discovered that it proceeded much more quickly and smoothly than it could have without the pre-work.

Samsung faces difficult times ahead just like many other companies in the semiconductor and electronics business today. North Korea is posturing about its nuclear capabilities, SARS has decimated markets in China, and memory chips still sell at rock-bottom prices. So what is Samsung's response? It has announced bold plans to invest $5.65 billion this year, about two-thirds more than last year. This includes $500 million to expand and upgrade its chip plant in Austin, TX, to make advanced memories for high-end servers. Somehow we have the feeling that the money will be well spent.

Robert Haavind
Editor in Chief