Matrix arithmetic accelerators that combine electronic circuits and photonics are one way silicon slingers can achieve better levels of performance, according to Chinese academics, who have drawn attention to a growing corpus of research in this area.
Optoelectronic processors, according to academics Pengfei Xu and Zhiping Zhou, might not be appropriate for all general-purpose computations. Optoelectronic chips, however, have the potential to surpass the computation performance of conventional electronic processors in terms of energy efficiency, computational power, latency, and maintainability when it comes to performing matrix operations, which are essential for high-performance workloads like AI, simulations, and graphics rendering.
“We believe that silicon-based optoelectronics is a promising and comprehensive platform for general-purpose matrix computation in the post-Moore’s law era,” Xu and Zhou wrote. The researchers contend that optoelectronic chips could outperform conventional computer processors in three critical areas: faster data transfer, reduced latency, and reduced power consumption. That is to say, light can assist with quick communications in addition to matrix operations.
“Instead of power-hungry electronic transceiver circuits, on-chip optical transceivers are good alternatives for low-energy-budget interconnections and boosting the data movement among the processors, memory, and peripheral hardware,” the duo wrote. “By exploiting the advantages of light in linear matrix computations, the photonic core is excellent at disruptively improving the computing performance, while the electronic circuits are necessary for performing other non-linear operations, such as driver circuits, arithmetic and logic, data storage, and activation function.”
The two claim that because optoelectronic chips’ total energy consumption is “merely proportionate” to a matrix’s side length, they can be especially useful for large-scale matrix processing. This stands in stark contrast to electrical processors, where the energy usage rises as the number of matrix parts increases. The results suggest that, if Moore’s law really does come to a stop some day, there may come a day when at least some computers and servers employ optoelectronic chips to power demanding tasks and may even be better than conventional computer chips.
