Yesterday, I attended the Global Technology Conference, a chip-making extravaganza underwritten by GLOBALFOUNDRIES and its partners. Got several blog posts to write about this information-packed day but thought I’d start with the three lessons that GLOBALFOUNDRIES learned from its 28nm high-K, metal-gate development, as told by Gregg Bartlett, Senior VP of Technology and Engineering at GLOBALFOUNDRIES.
Here are the lessons:
- Materials integration is as important as materials selection.
- SRAM is not a good yield predictor for advanced application-processing chips at 28nm.
- HKMG (high-K, metal-gate) process technology is hard to perfect, but it can be tamed.
So what is GLOBALFOUNDRIES getting for its blood, sweat, and tears from 28nm?
- 100% density increase relative to 40nm.
- 40% speed increase and 40% switching energy reduction relative to 40nm.
- 50% speed increase relative to 40nm with overdrive.
- 10-20% smaller die relative to 28nm gate last due to routability