IBM, for example, takes out many, many more patents per year than Intel. But are those patents a measure of success?
The basic difference between the two companies and the way they approach research is Intel's focus on products, as compared with IBM's much broader approach.
IBM's research and development organisation focuses on basic research. It develops new ideas and technologies based on a general strategy. The overall strategy and set of guidelines are renewed once per year. Keeping in mind the guidelines, IBM researchers are basically free to follow their own path. Product groups, such as IBM's personal systems group, then pick up the technologies for use in products.
Intel, on the other hand, adopted a methodology that applies its research efforts toward a specific goal, such as meeting a product road map. Intel prefers to see basic research done by universities or industry consortia such as Sematech, company officials said. Once an idea is proven to be worthy, Intel researchers will then work to help build it as a stand-alone product or add it to a new generation of Intel products.
It's all about the approach. One project, the IBM WatchPad, came about through the interaction of IBM's personal systems group with IBM Research. The group came up with the idea based on customer demands. Then various elements of the device, such as the screen technology, were culled from IBM Research. The watch can display pictures, text and animation, as well as a person's schedule, by blocking off booked time on the watch face itself. IBM showed it off at PC Expo last month. It may come to market in the next three to four years, company officials said.
IBM officials are proud of their company's different approach.
"Every basic element of modern CMOS technology has resulted from IBM," said Bijan Davari, an IBM fellow and director of its Semiconductor Research and Development Centre (SDRC) in New York State.
The SRDC is responsible for moving ideas created by IBM Research into microprocessor production. It works to prove out an idea, develop it into a manufacturing process and, finally, conduct test manufacturing before the technology is deployed in IBM chip fabs.
If it were to develop a product such as the WatchPad, Intel would have got the idea from customers as well. However, the basic ideas for elements needed to build the device would likely have come from one of its multitude of university research projects. Intel sponsors several research programs, which account for hundreds of individual research projects at about 100 universities.
This approach has worked so far for Intel. But will Intel's applied method of research continue to support its needs? David Tennenhouse, Intel's director of research, said he believes Intel's current model will work effectively for another five to ten years.
"We're working on changes to make it more efficient and more effective," he said. "The basic question is: how do you harbour research that is going to make significant changes to one of those road maps? The other problem is: how do you apply research to a market with no road map?"
Also, "how do we create the ability to do experiments inside the company for items that are new to the company?" he asked.
Tennenhouse, formerly of the Massachusetts Institute of Technology and the US government's Darpa (Defence Advanced Research Projects Agency), where he was chief scientist, believes the answers lie in tighter relationship with universities.
Tighter relationships would foster a more efficient means of transforming ideas into products, he said. To increase the speed at which ideas are made into products, Tennenhouse said, Intel needs a way to jump on projects that look promising, providing more funding to them if necessary, in order to bring ideas into the company's own research labs more quickly.
See Chips Central for daily hardware news, including an interactive timeline of AMD and Intel's upcoming product launches.