So what if Moore's law ends?

I'm sure you've read all about it lately on various blogs etc. Already several months ago, it was AMD and also Broadcom's CTO telling engineers it was necessary to prepare for the end of Moore's law. More recently it was Robert Colwell who joined the discussion to say that Moore's Law [will be] Dead by 2022. An article by Brian Bailey asks if the end of Moore's law is really a threat for national security, as Colwell says it will be. The comments on this one are particularly interesting, and some are also a bit scary in my opinion. Some people seem to really believe that the end of Moore's law means that the USA will no longer have a technical advantage over all other countries in the world, and that scares them...

Anyway... so what if Moore's law ends? Well for software engineers, it means we won't get additional speed nor cores for free anymore; for hardware engineers, it means that the transistor budget will stop increasing all the time. I expect that this will encourage innovation, for instance:

  • coming up with new approaches to increase the number of transistors per chip, for instance with 3D chips,
  • reducing logic consumption by identifying logic that is common to several IP blocks (see the work of F. Palumbo concerning Multi-Purpose Systems),
  • redesign SoCs around these new constraints,
  • having hardware and software engineers design chips together.

This is a good motivation for us at Synflow, because we believe that Cx will make the last two points easier to achieve. In addition to making hardware design faster and easier, Cx also makes hardware design possible for software engineers. You could argue that High-Level Synthesis of C/C++ already does this, but if you had read my post about High-Level Synthesis, really, you wouldn't :-)

What do you think?