Electronics consume about 290 TeraWatt hours of electricity a year in the US, a small slice of the nation's estimated total annual consumption of about 3,700 TWh. Buildings draw the lion's share of the power, about 2,700 TWh a year, said Bruce Nordman, a researcher at Lawrence Berkeley National Lab in an opening talk.

Nordman called for engineers to create a radically new architecture for smart, networked buildings based on universal standards to curb power use. Today's building controls such as programmable thermostats can increase power waste because they are proprietary and hard to use, he said.

"These things need to be universal because the degree to which we embrace non-interoperable systems, we waste energy," he said.

The industry also needs to put more focus on residential power controls. Homes consume about 57 percent of in-building energy, compared to commercial buildings which use about 30 percent, he added.

At the systems level, Nordman is working on ways PCs and set-top boxes can store their state information in a network device so they can go into sleep mode while still being available to respond to network traffic. "This could save 50 percent of a system's power consumption which is well worth getting," he said.

In many cases, keeping a system in a sleep state as long as possible can save more power than reducing its active power consumption or turning a system off. "'Off' is a twentieth century idea," he said. "You get a small benefit from going to sleep to off, but a huge one going from an on state to sleep."

At the chip level, "it's time to consider a new quantum jump beyond CMOS, said Michael Fritze, a program director at the Defense Advanced Research Projects Agency.

Fritze detailed a handful of DARPA projects using materials such as carbon nanotubes and MEMS techniques to create chips that consume less power than today's CMOS devices. For instance, researchers such as Chenming Hu at Berkeley are developing band-to-band tunneling transistors that hold the promise of 25 times less active power and 100 times less standby power than CMOS circuits.

"It's a new kind of physics that uses tunneling to get a steeper [power up] slope," he said. "It's extremely difficult, but it's worth a shot."

Another DARPA project is using so-called spin-torque technique to create a non-volatile memory with the capacity with the capacity and response rates of DRAM and SRAM. It is one of at least two non-volatile memory efforts at the agency.

"None of us can predict which architecture could have the best potential--there are a lot of them being considered," he said.

Several startups are working on energy efficient alternatives to flash memory, some of which could hit the market in advance of the DARPA initiatives, said Bill Joy, a venture capitalist at Kleiner Perkins Caufield & Byers and a founder of Sun Microsystems who gave a keynote talk. In a video interview, Joy said startups are getting funding to work on a wide variety of on clean tech ideas, despite a tightening of purse strings in the VC sector.

Both financial and technical pressures are building for the next big leaps in energy efficient electronics. Intel researcher George Bourianoff described the five-fold reductions in circuit leakage the CPU giant got from its high-K metal gate 45 nm process now in production.

"The next factor of five will probably be harder to come by," he said. "We are making steady progress in all areas, but we will not be able to do that indefinitely, so we are looking at new materials and structures."

Bourianoff gave an overview of some of the most recent advances in areas such as spintronics and phase-change devices that may be needed in as little as four years as chip makers drive beyond CMOS, he said.

"People are being pushed out of the comfort zones and they are feeling pressure," he said. "The good news is options are narrowing down around materials such as graphene," he added.

The Berkeley Symposium on Energy Efficient Electronic Systems gathered about 100 researchers. It was sponsored by the Center for IT Research in the Interests of Society (CITRIS) that recently moved into a new facility at Berkeley that will house a nanofabrication lab.