Photonic processor could perform at the speed of light. George Washington University researchers have recently demonstrated the potential of photonic processors. In an experiment using a specially designed Photonic processor to run matrix multiplier equations in parallel. In other words, they want to see how many operations the processor could perform at the same time. The answer was the process and was able to perform 2 quadrillion every seconds. This is a performance level two or three magnitudes greater than a conventional process. The photonic processor achieves this staggering capability through the use of wavelength division multiplexing. In other words, the Photonic processor uses different wavelengths of light on a single optical fiber. This allows the processor to run multiple tasks at the same time without interference from each other.
Reason for Experiments
The drive to develop a photonic processor was to see if it could be a viable alternative to modern processors. Specifically, Google’s TPU chip which already out performs the more common GPUs. The primary use for both chips is in machine learn for which the two processes are capable of. However, both the GPU chip and the TPU chip have the same major limitation which is lack of parallel process. In other words, they can only perform one process which results in a slow and power-hungry process. In an attempt to find an alternative not bound by these limitations resulted in the photonic process.
The researchers state that the experiment shows that the only time limitation on the processor was the photons speed. That is to say, the processor carried out its function at the speed of light. It was noted that the chip had one drawback, it’s power consumption. The photonic processor chip consumes about 80 W of power, a substantial amount compared to TPU chips. However, this could be outweighed by its process speed. Specifically, if you could network several photonic processors into a single device. Such a device would no longer use cloud processing for large amount data. If a photonic processor installed in a smart and then linked to 5G network could be endless.
