Energy Requirements And Challenges For IoT Autonomous Intelligence At The Edge

Today’s computational landscape is vast and power hungry. Can it be sustainable?


Recently, on a cold February night at San Jose State University, I attended the fourth episode of the talk series IR4: The Cognitive Era. IR4 talks focus on the fourth Industrial Revolution that is currently taking place and how cognitive science affects education, careers, and life. This latest part involved four esteemed experts on cloud and edge computing and the ramifications of energy efficiency.

Today’s computational landscape is vast and power hungry. Massive data centers sit in remote locations tied to high-voltage transmission lines and thick data trunks that tie them in to the cyber. Inside their cavernous halls, racks upon racks of high-performance servers suck cool air over burning heatsinks to keep their silicon at tolerable temperatures. Elsewhere, billions of smartphones, smartwatches, and other tiny gadgets are repeatedly draining their lithium-polymer energy stores during the day and charging them back up at night.

The net effect of this insatiable hunger for energy is currently less than optimal, to say the least. Energy production often means carbon emissions, especially in developing parts of the world. All of this carbon is piling up in our atmosphere and trapping heat, resulting in rising temperatures around the globe. Rising sea levels and ferocious weather storms are sure to follow, scientists warn. So how do we avoid this apocalyptic future that looks like a bad Michael Bay movie?

Enter our brave heroes, the panelists of the latest entry of the talk series IR4: The Cognitive Era. They were led by the inimitable Jim Hogan, private investor in high-tech startups, former EDA executive, and staunch supporter of renewable energy.

First, there is Garima Thockchom, founder and CEO of Akasia. Akasia helps companies transition their private computing resources to the cloud. This increases overall efficiency by reducing maintenance overhead and machine idle time. Servers can be collected into huge, highly-automated clusters that are intelligently managed by competent system administrators and complex scheduling software. Machines can be assigned to work on any task from any company, and data is kept safe by strong encryption.

There are many opportunities to save power at the edge as well. Vojin Zivojnovic, co-founder and CEO of AGGIOS, unites hardware and software under new power management methodologies to reduce inefficiencies. Vic Kulkarni, VP and chief strategist of the Semiconductor Business Unit of ANSYS, promotes comprehensive simulation of power delivery networks across the chip, package, and system to reduce unnecessary design margins. He also lobbies for the breakdown of silos that pigeonhole engineering teams and produce sub-optimal silicon products. Drew Wingard, co-founder and CTO of Sonics, preaches the balance between computation and communication—raw sensor data at the edge must often be filtered down with local processing resources before being radioed to the mothership.

It may be hard to imagine that efforts to save microwatts in the next generation of chips can have any real impact when looking at a whole power grid. Over time, though, these advances add up to produce real gains. In households once dominated by massive appliances, regulations such as EnergyStar have greatly reduced the share of power consumed by refrigerators, washing machines, and the like to 40% of total energy usage. The remainder now mostly feeds consumer electronics. One notable gain in that area comes from Vojin’s work at AGGIOS, where a firmware update to domestic digital gateways was found to reduce power consumption by over 30%. When multiplied by the millions of gateways that exist in the world, the total savings are equivalent to half of a 500 MW coal-fired power plant.

Continuing this fast pace of improvement is going to require new and improved simulation tools. These utilities must be made more accessible and comprehensive to evaluate the real-world impact of decisions at all stages of the design process. There are many organizations that are unable to attain or unaware of the benefits of pervasive simulation. For the groups that do value it and already apply it heavily in their workflows, there are still elusive corner cases or large-scope issues that their tools cannot adequately cover. This gives ANSYS an opportunity to spread the word about their existing invaluable solutions while creating new ones to enable the power-efficient infrastructure of the future.

And check out the webinar  “System-Level Chip-Package-System Power Integrity Co-analysis Solutions For 3D-ICs”  to learn about ANSYS chip package system solutions for designing power efficient electronics systems.

Leave a Reply