We All Clenched Our Teeth
What is the reason for setting the operating frequency of Cell at 4GHz?
This is just as initially targeted. The reason for setting this target was quite simple. I extrapolated the operating frequency for 2005 from logarithmic graphs, and it turned out be 4GHz. It is true that there were some negative voices calling to give up this target in the course of development. But I pleaded with the members to hold up-and we did it! Why was this possible? Because we developed the technology all the way from scratch. It would not have been possible if we had hung on to legacy technology. SPE is a completely new architecture with a simplified structure, and so it was relatively easy to achieve the target. There arose problems in regard to the Power-based CPU core, too. Some members proposed to give up on compatibility with the past instruction set, but I insisted not to. It was the existence of an architect, one of the best in IBM, which brought us out of the stalemate. This engineer found a way to create a totally new Power core while preserving the existing instruction set.
Do you intend to sell Cell in the market?
We'll do so, and aggressively. Sony, Toshiba, and IBM will produce as many Cell chips as possible. I have no intension to let Cell be a closed product. This is the era of the Internet. All kinds of devices are connected to the network, and I want to see Cell embedded in all of these devices. I believe openness will be the key from now on. Sony is good at making home-use products, but we also welcome others to do what we are not well-suited to doing, for example, an outrageous simulation machine which uses liquid-cooled Cells in a chain of ten thousand or more operating at some 10GHz. With such a development, we may become able to get what now takes a year done in a day, a day's work in ten minutes and ten minutes work in a second.
So, Sony will be putting more emphasis on the semiconductor business.
Most certainly. We will invest more in this business. The ratio of semiconductors in electronic devices already exceeds 50% and may rise to 70 or 80% in the future. Even the mounting technology may become considered as a semiconductor post-process. The semiconductor will no doubt gain importance in future device design work. But I would prefer to say it is more like Sony is establishing a computer business rather than a semiconductor business, as using more semiconductors means computerizing. Cell will evoke the appearance of graphics LSIs or southbridge chips, something almost like a PC, and that is the kind of business we want to start. For instance, I would like to think of even a TV receiver not as a display device, but as a potential computer. In short, Cell is what lies at the core as more and more devices are being computerized.
Do you plan to launch a Cell based game console in the near future?
We wouldn't have submitted the paper to the academic society unless we had some degree of prospect for commercialization. It would have been disrespectful to the society. Let me put it this way. I'm here for this interview, because we have a substantial level of confidence.