A new photonic chip design reduces energy needed to compute with light. Image: MIT A research group at MIT claims to have developed a ‘photonic’ chip requiring less power to operate than a conventional CPU, but which could be more efficient. The chip uses light in place of electricity, and simulation tests indicate that it could
A new photonic chip design reduces energy needed to compute with light. Image: MIT
A research group at MIT claims to have developed a ‘photonic’ chip requiring less power to operate than a conventional CPU, but which could be more efficient.
The chip uses light in place of electricity, and simulation tests indicate that it could process massive optical neural networks 10 million times more efficiently than today’s electrical chip-based computers.
For several years, researchers have been working to develop novel optical neural networks that use light to accelerate computation.
However, as optical neural networks (or traditional neural networks) grow in size and complexity, they start consuming much more power.
To address the issue, researchers have designed some specialised electrical chips, called AI accelerators, that can enhance the efficiency and speed of testing and training neural networks. But, one limitation of such accelerators is that they consume too much power as they scale.
AI & Machine Learning Live is returning to London on 3rd July 2019. Hear from the Met Office’s Charles Ewen, AutoTrader lead data scientist Dr David Hoyle and the BBC’s Noriko Matsuoka, among many others. Attendance is free to qualifying IT leaders and senior IT pros, but places are limited, so reserve yours now.
The new photonic accelerator chip developed by MIT researchers is characterised by optical components that are more compact in size compared to earlier versions.
Moreover, their chip uses optical signal-processing techniques to reduce power consumption requirements. The energy efficient “opto-electronic” scheme used in the chip helps in encoding data with optical signals.
According to researchers, the new photonic accelerator is scalable to large networks and can be operated at high (gigahertz) speeds and very low (sub-attojoule) energies.
The team carried out simulated training of neural networks using a MNIST image-classification dataset, which suggested that the new photonic accelerators could theoretically process optical neural networks about 10 million times below the energy-consumption limit of electrical chip-based accelerators.
The researchers believe their novel photonic accelerator could be used in future to reduce energy consumption in data centres running neural networks.
The detailed findings of the study are published in journal Physical Review X.
Delta is a new market intelligence service from Computing to help CIOs and other IT decision makers make smarter purchasing decisions – decisions informed by the knowledge and experience of other CIOs and IT decision makers.
Delta is free from vendor sponsorship or influence of any kind, and is guided by a steering committee of well-known CIOs, such as Charles Ewen, Christina Scott, Steve Capper and Laura Meyer.
Ten crucial technology areas are already covered at launch, with more data appearing and more areas being covered every week. Sign-up here for your free trial of the Computing Delta website.