The Optical Network-on-Chip (oNOC) processor is revolutionising AI applications with photonic technology and consistent power-latency dynamics.
Lightelligence has introduced a big data interconnect approach with the debut of Hummingbird. This represents the world’s inaugural Optical Network-on-Chip (oNOC) processor, designed for domain-focused Artificial Intelligence (AI) tasks. The company claims to employ vertical stacking techniques, combining a photonic and electronic chip into a unified package. This integrated design is the communication backbone for data centres and various high-calibre applications. With oNOC, power consumption and latency remain broadly consistent irrespective of distance. This makes the technology a prime candidate for crafting novel and sturdier topologies that aren’t dependent on proximate communication. The oNOC system streamlines the alignment of workloads with hardware, granting more latitude in choosing the aptest topology for the computational objective.
Hummingbird’s waveguides transmit signals at the speed of light, enabling an all-to-all data broadcast network to each core on a 64-core AI processor. This gives a latency and power advantage over traditional digital interconnects. The oNOC technology enables previously unattainable interconnect topologies, surpassing the limitations of digital networks. Hummingbird’s electronic and photonic ICs are co-packaged and formatted into a Peripheral Component Interconnect Express (PCIe) design, primed for integration into standard industry servers. Paired with the Software Development Kit (SDK), the system can be fine-tuned for machine learning and AI workloads, maximising the potential of the oNOC.
“Photonics is the solution to the critical compute scaling problem, which has become pressing as the traditional solutions struggle to keep up with the exponential growth of compute power demand spurred by breakthroughs in the AI industry,” remarks Yichen Shen, CEO of Lightelligence. “Hummingbird demonstrates how the industry can address the scaling problem by incorporating photonic technologies into their next-generation product.”
For more information, click here.