Scientists have developed a completely new type of microchip that uses microwaves instead of traditional digital circuits to perform its operations.
The processor can run faster than traditional CPUs and is the world’s first fully functional microwave neural network (MNN) that can be mounted on a chip, scientists reported in a study published Aug. 14 in the journal Nature Electronics.
you may like
“Because it can be instantaneously distorted across a wide range of frequencies in a programmable way, it can be reused for several computing tasks,” study lead author Bal Gobind, a doctoral student at Cornell University, said in a statement. “It bypasses many of the signal processing steps that digital computers normally have to do.”
the power of microwaves
The chip uses analog waves in the microwave range of the electromagnetic spectrum within an artificial intelligence (AI) neural network to impart a comb-like pattern to the microwave waveform. The regularly spaced spectral lines in the frequency comb act like a ruler, allowing you to measure frequencies quickly and accurately.
The neural network behind the microwave chip is a collection of machine learning algorithms inspired by the structure of the human brain. Microwave brain chips use interconnected electromagnetic nodes in tunable waveguides to identify patterns in datasets and adapt to incoming information.
The microwave brain was created using an MNN, an integrated circuit that processes spectral components (individual frequencies within a signal) by capturing features of the input data over a wide bandwidth.
The chip can solve simple logical operations and advanced calculations, such as recognizing binary sequences and identifying patterns in high-speed data, with 88% accuracy. In the study, the scientists said they proved this through several radio signal classification challenges.
By operating in the microwave analog range and applying a stochastic approach, the chip can process data streams on the order of tens of gigahertz (at least 20 billion operations per second). This speed exceeds that of most consumer computer processors (2.5 to 4 billion operations per second), which typically operate at 2.5 to 4 GHz.
“Bal abandoned much of traditional circuit design to accomplish this,” co-senior author Alyssa Apsell, chair of electrical and computer engineering at Cornell University, said in a statement. “Rather than trying to exactly mimic the structure of a digital neural network, he created something like a hodgepodge of controlled frequency behaviors that could ultimately lead to high-performance computation.”
you may like
Traditional digital systems require more circuitry, more power and more error correction to maintain accuracy, Govind added in a statement. However, the probabilistic approach allowed researchers to maintain high accuracy in both simple and complex calculations without increasing overhead.
The low power consumption of microwave chips is also noteworthy. It consumes less than 200 milliwatts (less than 0.2 watts), which is about the same transmit power as a mobile phone. By comparison, most CPUs require at least 65W of input power.
This low power consumption means the chip could be found in personal devices and wearable technology, the scientists said. This is a promising technology for use in edge computing because it can reduce latency by eliminating the need to connect to a central server. It also has the potential to aid in AI adoption, as it can provide a high-processing alternative with low power requirements for training AI models.
The researchers’ next step is to simplify the design by reducing the number of waveguides and making the chip smaller. More compact chips can use interconnected combs, which produce a richer output spectrum and are useful for training neural networks.
Source link
