Configurable Thoughts

 

One Size Fits All?

For the automobile industry, the year 1908 was a landmark, with the introduction of the Model T. Henry Ford made automobiles affordable for the middle class in the US, soon to be followed by the rest of the world. The Model T was in production from 1908 – 1927, with more than 15 million cars sold.

The most commonly known innovation for the Model T was the use of an assembly line, versus hand-crafted autos by individual mechanics/engineers. With this new way of production the company was able to keep cost low (large consumer market) and ramp up volumes. Production could even be spread around the globe with manufacturing sites eventually in Europe, South America and Asia.

More interesting is Henry Ford’s marketing concept: “Any customer can have a car painted any color that he wants so long as it is black”: Henry Ford made this statement in 1909, only a year after production of the Model T started. What is little known is that it took him until 1914 to actually implement this. From 1908–1914 the T-Ford was available in grey, red, green and blue. In 1914 he managed to have this specification change (reduction in available options) implemented, and the Model T was finally only available in black.

Like cars semiconductors were once handcrafted pieces of art. Over time libraries of transistors were introduced (compared to each designer making his own), for the same reason that Ford created the assembly line (cheaper, repeatable quality, etc.). Later this was replaced by the standard cells, and following that with RTL design and 3rd party IP cores. I still recall senior design engineers complaining that they had lost their creative freedom. As they used the more advanced capabilities and IP they became much more productive, and could build large complex SoCs and microprocessors. They no longer needed to alter transistor parameters, change standard cells, manually create netlists, or design each IP core. But rather than a restriction they found that even though they felt that they had less flexibility they could actually do more.

The availability of subsystems will be the next productivity improvement in semiconductor design, with IP suppliers providing solutions that are not only pre-integrated and tested, but also optimized for the end application, providing more value. With the availability of subsystems focus will shift away from hardware to software and ease of integration into the application. SoC designers will be able to directly integrate complete hardware and software functions into their design saving them time, increasing quality and enabling them to build even bigger SoCs.

Of course, unlike the Model T, subsystems will have to be flexible and available with a range of options so that users can tailor them to best fit their application needs. Although Henry Ford was successful one size does not fit all, and IP suppliers that are able to provide flexible subsystem solutions will be best positioned to support their customers.