Many companies, including such giants as Intel and IBM, are developing the topic of optical computers built on the principles of photonics. But in advanced areas, the most interesting solutions are often offered by young companies. An example is the Photonic Arithmetic Computing Engine (PACE), developed by startup Lightelligence. It is said to be vastly superior to GPUs in some of the toughest math problems.

More specifically, PACE has been successfully used to calculate the Ising model used to understand phase transitions. On this task, the optical processor showed 100 times faster speed compared to the Nvidia GeForce RTX 3080. It outperformed even the FPGA-based system specially designed for the Ising model by Toshiba specialists in performance. In this case, the superiority was 25 times.

PACE integrates 12,000 optical instruments operating at 1 GHz. PACE uses standard silicon photonics integration with the Mach-Zehnder interferometer (MZI) for computation and MEMS for waveguide reshaping in the MZI.

According to the source, this example shows that optical computing is very promising for solving certain problems. These are computationally expensive classes of mathematical problems, such as the Ising model, that are common in materials science, thermodynamics, bioinformatics, cryptography, circuit design, power grid optimization, and other fields.

bbabo.Net