SAN FRANCISCO Speaking here at Semicon West Tuesday (July 17), keynoters from processor arch-rivals Intel and AMD explored differing aspects of what it will take to survive in the semiconductor industry going forward. While one called for cooperation for today, the other emphasized innovation for tomorrow.
A senior executive from Advanced Micro Devices called for the industry to "band together to maximize efficiency," while Intel's chief technology strategist sounded more like Spartacus leading his troops to battle at the next technology node.
Doug Grose, senior vice president for technology development, manufacturing and supply chain at AMD, took the company's founder to task by offering a caveat to Jerry Sanders' infamous stand from the 1980s, in which Sanders declared: "Real men have fabs." Grose added: "Real men have partners as well."
Meanwhile, Intel's Paolo Gargini sounded off on the feats of 60 years of achievements in the semiconductor industry and offered five lessons from that history summed up as the common belief that "Everything that had to be invented has been invented in the last 30 years." But to think it was an easy ride "is a mistake," he said.
Gargini, director of technology strategy for Intel and chairman of the International Technology Roadmap for Semiconductors (ITRS), exhorted his audience to think that more inventions are yet to come that will propel the industry into the next 10-year phase of its existence.
A full day of ITRS activities here Wednesday (July 18) should give hints as to what the road map will look like through 2020.
AMD's Grose noted that the market and technology forces in the semiconductor industry are forcing his company to look outside the industry for guidance on how to manufacture chips more efficiently. Noting that the cost of building a fab has risen 250 percent in the last 10 years, while the cost of developing a new microprocessor has gone up 400 percent in that period, Grose stated that the industry is working under a number of false assumptions. "For some reason, we in the industry think that the semiconductor industry is unique and that semiconductor manufacturing is such a complex process that it can't be streamlined," said Grose.
He said AMD's ATI acquisition last year has helped fill out AMD's product portfolio but has also shown that AMD needs to adjust to the new "high-volume, high-demand/low-volume, low-demand" conundrum of the consumer electronics era.
AMD has looked to automakers Toyota and Porsche for advice on improving productivity in its fab. "Toyota has cut its labor costs to yield a 29.63 hours-per-vehicle operation, and Porsche has cut its inventories by 90 percent, for example. There are lessons to be learned from those two efficiency innovators," he said.
For AMD, operational efficiency has been key at its fab in Dresden, Germany, where in 2006 wafer starts increased 31 percent while overall wafer costs were cut by 26 percent. Overall productivity at the fab was boosted by 77 percent, the executive said.
Grose also reported that at AMD's back-end IC packaging plant in Singapore, the company was able to achieve a 47 percent reduction in cycle time and an increase of 25 percent in space efficiency, "and we are nowhere near close to hitting the wall. In fact, we are going to jump over the [fab] wall."
"As an industry, we need to develop a value chain model for our customers, as opposed to running the usual supply chain model used by the semiconductor industry over the years," Grose said.
The next-generation factory may look like many of today's fabs on the outside, but inside the fab "will be characterized as an end-to-end proposition encapsulating process, design and the supply chain," he said.
Grose noted that the next-generation factory concept will be able to support 300-mm and 450-mm wafers.
AMD's partnerships with IBM and Chartered, and with research institutes CNT and Albany Nanotech Center, are proving out next-generation ICs and processes, Grose said. The work is enabling new techniques, such as immersion lithography, that will take the company into the 32-nm processing node, he said.
Intel's Gargini, meanwhile, extolled the virtues of ITRS as it celebrates its 10th year as the first international road map for any industry. He gave a historical perspective to past innovations in the semiconductor industry that "today we take for granted, but that were very hard to achieve."
"Every time engineers came up to a wall of obstacles, they were able to find a solution to go beyond," said Gargini. "Don't think for a minute that these achievements were obvious at the time of their birth, but that's the nature of the creative process. We pay engineers good money to tackle the next problem."
Gargini said young engineers entering the business today may think from observations that the semiconductor industry is approaching the limits of its physics, "but there is an entire subatomic world we have not explored yet."
He said three large research consortiums have sprung up on the both U.S. coasts and in the middle of the country. "Intel has taken the lead in the West, IBM in the East, and Texas Instruments is leading the Austin research activities in carbon nanotubes, spin electronics and advanced material growth," he said.
"Cooperation is paramount, and innovation is critical as well," said Gargini.