IBM research takes broader role

Electronic News, April 21, 1997 by Gale Bradley

You're liable to miss hearing before the summer where the silicon germanium work is (see The Antenna, page 4), or later what the relatively new Austin, Texas, 30-person microprocessor research team is doing, or even the manufacturing process innovations--akin to those already unleashed, like wafer polishing and laser memory repair--that IBM has in store.

And this begs the question: with all this in the pipeline, aren't the powers that be at corporate IBM occasionally raising the stick a bit, asking that IBM, and not Intel or Motorola, make more money from these microelectronic innovations?

"Every day," John Kelly, IBM Microelectronics' VP for strategy and technology, tells Electronic News. "Actually, every day, we are asking that of ourselves. That's what we spend a great deal of our time doing."

"We're a very technology-rich company and we've learned over the last five, maybe four, years to couple our research and process development efforts into joint projects," he said.

And that IBM has done, first on a smaller scale with the 64-megabit DRAM project, headed by Randy Isaac, in conjunction with Siemens and later Toshiba, and now on a large scale with the team of Microelectronics and Research silicon germanium (SiGe) engineers under Research's Bernard Meyerson.

"Now we need better and better coupling with our customers," Mr. Kelly said.

One result of that recognition is IBM Micro's new Integrated Platform Solutions group, headed by VP Michael P. Concannon, and, in fact, an overall platforms supplier market orientation. The first shot went off with their set-top box reference design unveiled just before the National Association of Broadcasters (NAB) exhibition earlier this month in Las Vegas (EN, March 31).

The approach is one Lucent Technologies' Microelectronics group started taking last fall with its Silicon Suite offerings, now numbering three.

IBM Microprocessors

Down in Texas, the 18-month-old Austin Research Lab in now populated with some 30 microprocessor engineers under director Michael Blasgen with one goal: bringing the IBM AS/400 and RS/6000 computers a 1GHz CPU, according to Mr. Blasgen.

"We have chosen a thousand megahertz, a gigahertz, as our goal. That's five times what you can buy today, and it's a rather difficult challenge," Mr. Blasgen said.

IBM was not ready to talk much more about the leading edge of the microprocessor work in Austin, but various executives did say Karl Anderson, who couldn't be reached by press time, formulated a team last fall that drew talent from Watson's basic physical sciences arsenal, PowerPC developers and semiconductor architects from outside of IBM.

"I can't tell you any more about (Mr. Anderson's work)," Mr. Blasgen said. "Not only because it's a secret, but also because it's a few years away. It's not going to hit the streets 14 months from now." Still, he cautioned not to get caught up in "Super CPU" delusions of grandeur. "It's going to be based on the PowerPC core set of instructions. We're not going to start a whole new thing here."

How will they get to the 1GHz CPU? "It's not enough to use the latest silicon technology," Mr. Blasgen said. Does that mean the semiconductor material or processing approach will change to get there?

"No, we're working in conventional CMOS, that's not to say conventional but the accepted, leading-edge CMOS process," Mr. Blasgen explained. "We're not a process laboratory, most of our expertise (in process) is still at the Watson lab and in Fishkill.

"They have a long history (of bringing forth the next processing steps) and that's very important to us. It sort of cost an extra couple hundred million dollars to be able to do that, but we do, and we're one of only a very few who do," he said.

Fab Line Innovations

Mr. Kelly corroborated Mr. Blasgen's point. "If you look at a lot of the technology that's now standard in the industry. Chemical mechanical polishing, shallow trench isolation, tungsten interconnects...All of these technologies came out of these joint (IBM Research/IBM Microelectronics) projects. If you look at what's being done at Intel, at Motorola. Those roots were from these (IBM-wide development programs)," Mr. Kelly said.

And there are indications, from different quadrants, that the T.J. Watson Research Center is studying semiconductor fabrication as closely as ever. Last Monday, IBM announced that I.C. Hoyan of Watson and researchers at Columbia University are working together in the next two to three years to measure the forces behind the electron migration that can occur in the aluminum layer of a CMOS device.

Whether this new phase of development kicked in because IBM needs to measure those electron migration forces in copper layers, Mr. Hoyan chose not to say. In fact, certain Sematech engineers were meeting at the Brookhaven National Labs in Long Island, N.Y., last Wednesday to see the extremely powerful atomic force microscope Mr. Hoyan sometimes uses for his work.

Then on Tuesday in San Francisco, the American Chemical Society honored Watson materials chemist Christopher Murray for some of his research with MIT's Moungi Bawendi on cadmium selenide semiconductors. This was most notable for how Mr. Murray described the project: