What to do with 1 billion transistors?
In the thirty years that I’ve been in the semiconductor business we’ve all known about Moore’s law. And we’ve all used it to project out a few years to rationalize product plans. But sometimes the implications of a trend aren’t fully grasped.
Intel recently announced that they are sampling a 1 billion transistor member of their processor family – the Montecito processor. But that’s not the main news.
In 1990 the DSP product line that I was responsible for at Analog Devices. The main processor contained about 100k transistors. The mixed signal baseband processor just two years later was about 250k transistors. In those days we paid close attention to layout optimization, circuit choice, and multiple uses of any circuit element we could reasonably use. Over the years architectures have been developed to exploit the increased number of transistors that could be put on a single die.
There’s about to be fundamental changes in audio devices fueled by the exponential increase in the number of new transistors possible per die from semiconductor process generation succession:
Audio will become a part of products as opposed to being the main focus.
Audio codecs will increasingly be developed to execute on DSP-like processors.
DSP-like processors will be duplicated on a single die to bring more processing power to consumer applications.
Memory sizes will permit significant buffering of streaming data of all types.
Consumer ICs will become memories with processing distributed throughout the memory array.
Many of these changes will be required unless the industry can develop new semiconductor processing techniques to process a single wafer at a time in a cost competitive manner.
The question for audio developers is what steps will you take to adapt to the coming change?
I’d like to hear your comments.