via Stanford University
Today’s electronics waste energy by using separate chips to process and store data. New algorithms enable energy-efficient hybrids to work together as one mega-chip.
Smartwatches and other battery-powered electronics would be even smarter if they could run AI algorithms.
But efforts to build AI-capable chips for mobile devices have so far hit a wall — the so-called “memory wall” that separates data processing and memory chips that must work together to meet the massive and continually growing computational demands imposed by AI.
“Transactions between processors and memory can consume 95 percent of the energy needed to do machine learning and AI, and that severely limits battery life,” said computer scientist Subhasish Mitra, senior author of a new study published in Nature Electronics.
Now, a team that includes Stanford School of Engineering computer scientist Mary Wootters and electrical engineer H.-S. Philip Wong has designed a system that can run AI tasks faster, and with less energy, by harnessing eight hybrid chips, each with its own data processor built right next to its own memory storage.
This paper builds on the team’s prior development of a new memory technology, called RRAM, that stores data even when power is switched off — like flash memory — only faster and more energy efficiently. Their RRAM advance enabled the Stanford researchers to develop an earlier generation of hybrid chips that worked alone. Their latest design incorporates a critical new element: algorithms that meld the eight, separate hybrid chips into one energy-efficient AI-processing engine.
“If we could have built one massive, conventional chip with all the processing and memory needed, we’d have done so, but the amount of data it takes to solve AI problems makes that a dream,” Mitra said. “Instead, we trick the hybrids into thinking they’re one chip, which is why we call this the Illusion System.”
The researchers developed Illusion as part of the Electronics Resurgence Initiative (ERI), a $1.5 billion program sponsored by the Defense Advanced Research Projects Agency. DARPA, which helped spawn the internet more than 50 years ago, is supporting research investigating workarounds to Moore’s Law, which has driven electronic advances by shrinking transistors. But transistors can’t keep shrinking forever.
“To surpass the limits of conventional electronics, we’ll need new hardware technologies and new ideas about how to use them,” Wootters said.
The Stanford-led team built and tested its prototype with help from collaborators at the French research institute CEA-Leti and at Nanyang Technological University in Singapore. The team’s eight-chip system is just the beginning. In simulations, the researchers showed how systems with 64 hybrid chips could run AI applications seven times faster than current processors, using one-seventh as much energy.
Such capabilities could one day enable Illusion Systems to become the brains of augmented and virtual reality glasses that would use deep neural networks to learn by spotting objects and people in the environment, and provide wearers with contextual information — imagine an AR/VR system to help birdwatchers identify unknown specimens.
Stanford graduate student Robert Radway, who is first author of the Nature Electronics study, said the team also developed new algorithms to recompile existing AI programs, written for today’s processors, to run on the new multi-chip systems. Collaborators from Facebook helped the team test AI programs that validated their efforts. Next steps include increasing the processing and memory capabilities of individual hybrid chips and demonstrating how to mass produce them cheaply.
“The fact that our fabricated prototype is working as we expected suggests we’re on the right track,” said Wong, who believes Illusion Systems could be ready for marketability within three to five years.
The Latest Updates from Bing News & Google News
Go deeper with Bing News on:
- Unbricking A BluRay Driveon January 20, 2021 at 4:00 pm
All BluRay player, devices, and drives contain a key that unlocks the encryption and DRM present on BluRay discs. Since 2007, the consortium responsible for this DRM scheme has been pushing ...
- Stanford researchers combine processors and memory on multiple hybrid chips to run AI on battery-powered smart deviceson January 12, 2021 at 5:31 am
January 11, 2021 Stanford researchers combine processors and memory on multiple hybrid chips to run AI on battery-powered smart devices In traditional electronics, separate chips process and store ...
- Team creates hybrid chips with processors and memory to run AI on battery-powered deviceson January 11, 2021 at 8:04 am
Smartwatches and other battery-powered electronics would be even smarter if they could run AI algorithms. But efforts to build AI-capable chips for mobile devices have so far hit a wall - the so ...
- Marvell to buy Inphi, creating $40B semiconductor gianton October 29, 2020 at 7:17 am
Marvell Technology Group has agreed to buy Inphi Corp. for about $10 billion in cash and stock, making for the second mega chip deal this week. Continue Reading Below The chipmaker plans to ...
Go deeper with Google Headlines on:
Go deeper with Bing News on:
- French Researchers Exploit Non-Idealities of Memristors to Bring AI to the Edgeon January 23, 2021 at 2:17 pm
Keeping with the trend to depart from von Neumann architectures, French researchers have overcome non-ideal properties of memristors to bring Bayesian networks to the edge.
- KBRA Assigns Preliminary Ratings to RRE 6 Loan Management DACon January 23, 2021 at 12:07 am
Kroll Bond Rating Agency UK Limited (KBRA) assigns preliminary ratings to five classes of notes to be issued by RRE 6 Loan Management DAC, a cash flow collateralised loan obligation (CLO) backed ...
- Leti researchers use RRAM for MLon January 19, 2021 at 10:11 pm
CEA-Leti scientists have demonstrated a machine-learning technique exploiting what have been previously considered as “non-ideal” traits of resistive-RAM (RRAM) devices, overcoming barriers to ...
- Edge learning breakthroughon January 19, 2021 at 4:24 pm
In a paper published in the January issue of Nature Electronics, the research team demonstrated how RRAM, or memristor, technology can be used to create intelligent systems that learn locally at the ...
- CEA-Leti Reports Machine-Learning Breakthrough That Opens Way to Edge Learningon January 19, 2021 at 10:10 am
CEA-Leti scientists have demonstrated a machine-learning technique exploiting what have been previously considered as “non-ideal” traits of resistive-RAM (RRAM) ...