Unleashing a new horizon in the world of Artificial Intelligence (AI) applications, Intel unveiled the curtains on its latest revolutionary achievement of computational power and energy. During the Optical Fiber Communication Conference (OFC) 2024, Intel introduced the world its Optical Compute Interconnect (OCI) chiplet, the first of a kind fully integrated OCI to use light instead of electricity in data transmission between processors.
Intel introduced the chiplet through a running live data that demonstrated the blink of an eye process in a co-packaged Intel CPU.
The OCI chiplet comes providing solutions to the most-concerning data transmission process, which usually happens through a connection of electric signals traveling on copper cables.
Currently, data passes between memory modules and GPUs or CPUs through conventional copper interconnects. As the need of data transfer -with the use of AI- grows, comes also massive consumption of power.
In the introduced chiplet, with the use of laser light, transferring data can be done through fiber optic connections, making it possible for data to travel 100 longer than the typical way. The chiplet is also designed to support 64 channels of 32 gigabits per second (Gbps) data transmission in each direction.
This reduces power consumption while speeding the reach limit at the same time.
“The OCI chiplet boosts bandwidth, reduces power consumption and increases reach, enabling machine learn workload acceleration that promises to revolutionize high-performance AI infrastructure,” said Thomas Liljeberg, Intel’s senior director, Product Management and Strategy, Integrated Photonics Solutions Group in a press preview.
Intel’s groundbreaking achievement empowers customers to seamlessly integrate co-packaged silicon photonics interconnect solutions into next-generation compute systems, he added.
According to Intel, the revolutionary achievement represents a leap forward in high-bandwidth interconnect by enabling co-packaged optical input/output (I/O) in emerging AI infrastructure for data centers and high-performance computing (HPC) applications.
Liljeberg explained that the prototype is considered a “compact, fully self-contained I/O system” that combines two co-packaged chips, which are the photonic integrated circuit (PIC), and the electrical integrated circuit (EIC).
The newly introduced chip is expected the change the AI industry. The increasing need of AI applications, especially large language models (LLM) like Open AI’s ChatGPT, has also been reflected on the market needs of a more powerful and efficient chips. This requires playing a key role in addressing the emerging requirements of AI acceleration workloads.
“The ever-increasing movement of data from server to server is straining the capabilities of today’s data center infrastructure, and current solutions are rapidly approaching the practical limits of electrical Input/Output performance,” Liljeberg explained elaborating on the need of the OCI chiplet.
Intel’s next step is kicking off the delivery of the chiplet in large volumes to start integrating in in AI used applications. It further aims to trun pilot programs with various customers to start integrating the technology as production scales up.
