Intel Corporation has achieved a significant milestone in integrated photonics technology. At the Optical Fiber Communication Conference (OFC) 2024, Intel's Integrated Photonics Solutions (IPS) Group demonstrated the industry's first fully integrated optical compute interconnect (OCI) chiplet co-packaged with an Intel CPU and running live data.
This OCI chiplet represents a leap forward in high-bandwidth interconnect, enabling co-packaged optical input/output (I/O) in emerging AI infrastructure for data centers and high-performance computing applications.
The OCI chiplet is designed to support 64 channels of 32 gigabits per second (Gbps) data transmission in each direction on up to 100 meters of fiber optics. The OCI chiplet addresses AI infrastructure's growing demands for higher bandwidth, lower power consumption, and longer reach, making it a crucial advancement for future computing platforms.
OCI Chiplets Vs Traditional Electrical Interconnects
 |
OCI chiplets offer significantly higher bandwidth compared to electrical interconnects. With 64 channels of 32 Gbps data transmission in each direction, they provide ample capacity for data-intensive workloads. Traditional electrical interconnects, such as copper-based traces on circuit boards, have limited bandwidth and can become bottlenecks in high-performance computing systems.
Moreover, OCI chiplets consume less power per bit transmitted. Optical signals experience minimal resistance and don't generate heat like electrical currents do. Electrical interconnects suffer from power losses due to resistance, leading to higher energy consumption.
OCI chiplets support longer reach—up to 100 meters of fiber optics—making them suitable for large-scale data centers. Electrical interconnects are limited by signal degradation over distance, especially at high speeds.
Optical signals travel at the speed of light, resulting in lower latency compared to electrical signals. Electrical interconnects introduce additional latency due to signal propagation delays.
Above all, OCI chiplets are immune to EMI, making them ideal for noisy environments. While, Electrical interconnects can suffer from EMI-induced signal degradation.
In summary, OCI chiplets offer superior performance in terms of bandwidth, power efficiency, reach, and latency, making them a promising solution for future computing systems.
OCI chiplets for AI-based Applications
With recent developments in large language models (LLM) and generative AI are accelerating the trend of AI applications. Larger and more efficient machine learning (ML) models will play a key role in addressing the emerging requirements of AI acceleration workloads. The need to scale future computing platforms for AI is driving exponential growth in I/O bandwidth and longer reach to support larger processing unit (CPU/GPU/IPU) clusters and architectures with more efficient resource utilization, such as xPU disaggregation and memory pooling.Electrical I/O (i.e., copper trace connectivity) supports high bandwidth density and low power, but only offers short reaches of about one meter or less. Pluggable optical transceiver modules used in data centers and early AI clusters can increase reach at cost and power levels that are not sustainable with the scaling requirements of AI workloads. A co-packaged xPU optical I/O solution can support higher bandwidths with improved power efficiency, low latency and longer reach – exactly what AI/ML infrastructure scaling requires.
Advertisements