Chipmakers pursue new architectures to address deep learning

Jan. 12, 2017

The semiconductor industry is pursuing new architectures as Moore’s Law begins offering diminishing returns. Options highlighted last November at the International Test Conference ranged from heterogeneous integration to FPGA-based hardware inference accelerators.

Now in The Wall Street Journal Don Clark notes that the pressures facing the $335 billion global semiconductor industry “…are spurring a renaissance of semiconductor innovation, along with a growing band of startups aiming to exploit it….”

Part of the pressure on the industry stems from a slowdown in cellphone and PC business and consolidation in the industry. He cites Dealogic tallies that 707 mergers and acquisitions over the past two years have totaled $246 million.

Big and small companies, Clark says, are working on new chip designs, materials, and manufacturing processes as they focus on deep learning as it applies to image classification, speech translation, and autonomous driving.

“The Semiconductor Industry Association and its research affiliate have enlisted 22 tech companies to launch a broad study of technologies that might bring computing advances,” Clark writes. “Alternatives range from stacking circuitry in space-saving layers to making chips from biological materials such as proteins.”

Among the highlights:

  • Alphabet’s Google is designing a chip from scratch for deep learning.
  • IBM’s TrueNorth includes a million structures patterned after the brain’s neurons.
  • Startups including Cerebras Systems, KnuEdge Inc., Graphcore Ltd., Cornami, and Wave Computing are all addressing deep learning.
  • Microsoft is augmenting the servers in its Azure cloud service with FPGAs to accelerate chores such as Bing searches.

Clark quotes Doug Burger, a distinguished engineer at Microsoft, as saying, “I need more than what Intel and AMD are doing in CPUs. That is why we have this Cambrian explosion of new processor architectures.”

Intel is not standing still. For example, it acquired FPGA maker Altera in 2015. Clark concludes his article by quoting Jason Waxman, vice president and general manager of Intel’s data center group, as saying, “I’m excited to see a lot of innovation and ideas. I tend to believe very few of them are going to last as discrete products.”

About the Author

Rick Nelson | Contributing Editor

Rick is currently Contributing Technical Editor. He was Executive Editor for EE in 2011-2018. Previously he served on several publications, including EDN and Vision Systems Design, and has received awards for signed editorials from the American Society of Business Publication Editors. He began as a design engineer at General Electric and Litton Industries and earned a BSEE degree from Penn State.

Sponsored Recommendations

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!