June 20, 2019 at 19:15Reduce energy consumption in the data center - a new photon chip will help
MIT has developed the architecture of the new photon processor. It will increase the efficiency of optical neural networks by a thousand times, compared with similar devices.
The chip will reduce the amount of electricity consumed by the data center. We tell how it is arranged. Photos - Ildefonso Polo - Unsplash
Optical neural networks are faster than traditional solutions using electronic components. Light does not require isolation of signal paths, and laser streams are able to pass through each other without mutual influence. Thus, all signaling paths can function simultaneously, which ensures a high data rate.
But there is a problem - the larger the size of the neural network, the more energy it consumes. To solve this problem, special accelerator chips (AI accelerators) are being developed that optimize data transfer. However, they do not scale as well as we would like.
The solution to the problem of energy efficiency and scaling of optical chips took up MIT and introducedThe new architecture of the photon accelerator, which reduces the power consumption of the device by a thousand times and works with tens of millions of neurons. The developers say that in the future the technology will find application in data centers that interact with complex intelligent systems and machine learning algorithms, as well as analyze big data.
The new chip is based on an optoelectronic circuit. The transmitted data is still encoded with optical signals, but balanced homodyne detection is used for matrix multiplication ( p. 30 ). This is a technique that allows you to form an electrical signal based on two optical.
To transmit light pulses with information about the input and output neurons, one signal path is used. Data on the weights of neurons, on the contrary, come through separate channels. All of them “diverge” along the grid nodes of homodyne photodetectors, which calculate the output value for each neuron (determine the signal level). This information then goes to a modulator, which converts the electrical signal back to optical. Next, it goes to the next layer of the neural network and the process repeats.
In their scientific work, engineers from MIT give the following scheme for a single layer:
Image: Large-Scale Optical Neural Networks Based on Photoelectric Multiplication / CC BY The
new architecture of the AI accelerator requires only one input and one output channel for each neuron. As a result, the number of photodetectors is equal to the number of neurons, and not their weights.
A similar technology is being developed by Lightelligence, a small startup from Boston. Employees of the company say that their AI accelerator will allow solving machine learning tasks hundreds of times faster than classical devices. Last year, the team completed the prototype of their device and was preparing to conduct tests.
Works in the field of photonic chips and Cisco. At the beginning of the year the company announced a purchaseLuxtera startup, which designs photonic chips for data centers. In particular, the company produces hardware interfaces that allow you to connect fiber directly to servers. This approach increases network bandwidth and speeds up data transfer. Luxtera devices use special lasers for encoding information and germanium photo detectors for its decryption.
Photos - Thomas Jensen - Unsplash
Other major IT companies such as Intel are also involved in optical technology. Back in 2016, they began to produce their own optical chips that optimize data transfer between data centers. Recently, representatives of the organization toldthat they plan to implement these technologies outside of data centers - in lidars for unmanned vehicles.
So far, photonic technology cannot be called a universal solution. Their implementation requires high costs for the technical re-equipment of data centers. But developments like those being developed at MIT and other organizations will make optical chips cheaper and will most likely allow them to be “promoted” in the mass market for data center equipment.
We at ITGLOBAL.COM help companies develop their IT infrastructure and provide private and hybrid cloud services. Here's what we write about in our corporate blog:
The chip will reduce the amount of electricity consumed by the data center. We tell how it is arranged. Photos - Ildefonso Polo - Unsplash
Why do we need a new architecture
Optical neural networks are faster than traditional solutions using electronic components. Light does not require isolation of signal paths, and laser streams are able to pass through each other without mutual influence. Thus, all signaling paths can function simultaneously, which ensures a high data rate.
But there is a problem - the larger the size of the neural network, the more energy it consumes. To solve this problem, special accelerator chips (AI accelerators) are being developed that optimize data transfer. However, they do not scale as well as we would like.
The solution to the problem of energy efficiency and scaling of optical chips took up MIT and introducedThe new architecture of the photon accelerator, which reduces the power consumption of the device by a thousand times and works with tens of millions of neurons. The developers say that in the future the technology will find application in data centers that interact with complex intelligent systems and machine learning algorithms, as well as analyze big data.
What is she
The new chip is based on an optoelectronic circuit. The transmitted data is still encoded with optical signals, but balanced homodyne detection is used for matrix multiplication ( p. 30 ). This is a technique that allows you to form an electrical signal based on two optical.
To transmit light pulses with information about the input and output neurons, one signal path is used. Data on the weights of neurons, on the contrary, come through separate channels. All of them “diverge” along the grid nodes of homodyne photodetectors, which calculate the output value for each neuron (determine the signal level). This information then goes to a modulator, which converts the electrical signal back to optical. Next, it goes to the next layer of the neural network and the process repeats.
In their scientific work, engineers from MIT give the following scheme for a single layer:
new architecture of the AI accelerator requires only one input and one output channel for each neuron. As a result, the number of photodetectors is equal to the number of neurons, and not their weights.
Such a trip allows you to save space on the chip, increase the number of useful signaling paths and optimize power consumption. Now engineers from MIT are creating a prototype that will test the capabilities of the new architecture in practice.
Who else is developing photonic chips
A similar technology is being developed by Lightelligence, a small startup from Boston. Employees of the company say that their AI accelerator will allow solving machine learning tasks hundreds of times faster than classical devices. Last year, the team completed the prototype of their device and was preparing to conduct tests.
Works in the field of photonic chips and Cisco. At the beginning of the year the company announced a purchaseLuxtera startup, which designs photonic chips for data centers. In particular, the company produces hardware interfaces that allow you to connect fiber directly to servers. This approach increases network bandwidth and speeds up data transfer. Luxtera devices use special lasers for encoding information and germanium photo detectors for its decryption.
Photos - Thomas Jensen - Unsplash
Other major IT companies such as Intel are also involved in optical technology. Back in 2016, they began to produce their own optical chips that optimize data transfer between data centers. Recently, representatives of the organization toldthat they plan to implement these technologies outside of data centers - in lidars for unmanned vehicles.
What is the result
So far, photonic technology cannot be called a universal solution. Their implementation requires high costs for the technical re-equipment of data centers. But developments like those being developed at MIT and other organizations will make optical chips cheaper and will most likely allow them to be “promoted” in the mass market for data center equipment.
We at ITGLOBAL.COM help companies develop their IT infrastructure and provide private and hybrid cloud services. Here's what we write about in our corporate blog: