Cell SupercomputerA Cell Industries Company
HomeAbout UsOpinion PollsMailing ListContact UsLinks
How can one chip be so fast?
Will the Cell concept work?
How will the Cell processor change the world?
Supercomputer History
Transputer
Other High Speed Chips and Efforts
IBM and PC History
IBM and World Economy
Cell Applications
Latest News
Information Disclosed at ISSCC
Power PC, Cell processor, Open source and You
Will the cell concept work?
STI got together in 2001 and said extreme processor complexity does not make any sense, so they went back to the drawing board and started to figure out how to put a large number of simpler cpu's on a single chip. The difficult part of the design effort was to allow all those cpu's to cooperate on one problem.

I think there is good evidence that several hundred up to 10,000 cpu's can work together economically. The evidence comes from Clusters of off the shelf PC's which now make up most of the top 500 supercomputers, All graphic processors (GPU's) use several pipelines, Playstation 2 uses uses mulitple vector units.

How well will the Cell work? My opinion is the X86 architecture was never the best, but it won in the market place. We will only ever know if the cell processor concept is really good if large numbers of people adopt it and write software for it. I think most technical people and organizations are at least open to a Wintel alternative and many would go out of their way to find one. If STI makes an honest effort to make everyone feel they are part of a revolution, the cell concept will catch on, it will work, and the cell will replace the X86.

I have looked at the patent. I think the design looks workable and I for one would like to get my hands on one a see what I could make it do.


United States Patent 6,526,491
Suzuoki ,   et al. February 25, 2003


Inventors: Suzuoki; Masakazu (Tokyo, JP); Yamazaki; Takeshi (Tokyo, JP)
Assignee: Sony Corporation Entertainment Inc. (JP)
Appl. No.: 816020
Filed: March 22, 2001