Intel has developed a new chiplet called the Optical Compute Interconnect (OCI) that utilizes light instead of electricity to enhance data transmission bandwidth. This chiplet is specifically designed to address the increasing demand for computational power and energy in AI applications. By integrating the OCI chiplet with its CPUs, Intel aims to enable processors to communicate over longer distances using optical signaling, which can be nearly 100 times longer than traditional copper connections.
The OCI chiplet combines optical connections with integrated circuits for processing information, electrical circuits for handling electrical signals, and optical connectors for light signal ports. This modular design allows for scalability and faster data transmission speeds between processors. While other companies like Global Foundries and Ayar Labs are also exploring photonics technology for semiconductors, Intel’s OCI chiplet is positioned to be compatible with various processors, including those from competitors like Nvidia and AMD.
Despite the potentially higher costs associated with optical technologies compared to conventional methods, Intel believes that the benefits and use cases of the OCI chiplet justify the price. The company is working on integrating this technology into processing platforms in large volumes and collaborating with partners to ensure interoperability with other photonic chips.
In the future, Intel plans to scale up production of the OCI chiplet to meet the increasing demand for data center processing capabilities. As more data centers adopt this technology, the chiplet is expected to support bandwidth requirements for future generations. With its focus on delivering cost-effective and scalable solutions, Intel aims to revolutionize data transmission in AI applications by introducing the OCI chiplet to the broader market.
Article Source
https://www.forbes.com/sites/kirkogunrinde/2024/06/26/intel-is-using-lasers-to-help-meet-ai-demands-on-data-centers/