What’s New: Intel Corporation has achieved a groundbreaking milestone in integrated photonics technology for high-speed data transmission. At the Optical Fiber Communication Conference (OFC) 2024, Intel’s Integrated Photonics Solutions (IPS) Group showcased the industry’s most advanced and first-ever fully integrated optical compute interconnect (OCI) chiplet. This OCI chiplet, co-packaged with an Intel CPU and running live data, marks a significant advancement in high-bandwidth interconnects, enabling co-packaged optical input/output (I/O) in emerging AI infrastructure for data centers and high-performance computing (HPC) applications.
“The ever-increasing data movement from server to server is straining today’s data center infrastructure, and current solutions are nearing the practical limits of electrical I/O performance. Intel’s groundbreaking OCI chiplet allows customers to seamlessly integrate co-packaged silicon photonics interconnect solutions into next-generation compute systems. Our OCI chiplet enhances bandwidth, reduces power consumption, and extends reach, enabling ML workload acceleration that promises to revolutionize high-performance AI infrastructure.”
–Thomas Liljeberg, Senior Director, Product Management and Strategy, Integrated Photonics Solutions Group
What It Does: This first OCI chiplet supports 64 channels of 32 gigabits per second (Gbps) data transmission in each direction over up to 100 meters of fiber optics. It addresses the growing demands for higher bandwidth, lower power consumption, and longer reach in AI infrastructure. It enables future scalability of CPU/GPU cluster connectivity and novel compute architectures, including coherent memory expansion and resource disaggregation.
Why It Matters: The deployment of AI-based applications is accelerating globally, driven by advancements in large language models (LLM) and generative AI. Larger and more efficient machine learning (ML) models are crucial for meeting the emerging requirements of AI acceleration workloads. Scaling future computing platforms for AI necessitates exponential growth in I/O bandwidth and longer reach to support larger processing unit (CPU/GPU/IPU) clusters and architectures, such as xPU disaggregation and memory pooling.
Electrical I/O, while offering high bandwidth density and low power, is limited to short reaches of about one meter or less. Pluggable optical transceiver modules used in data centers and early AI clusters can increase reach but come at unsustainable cost and power levels. A co-packaged xPU optical I/O solution supports higher bandwidths with improved power efficiency, low latency, and longer reach—essential for AI/ML infrastructure scaling.
How It Works: The fully integrated OCI chiplet leverages Intel’s field-proven silicon photonics technology, integrating a silicon photonics integrated circuit (PIC) with on-chip lasers and optical amplifiers, alongside an electrical IC. The OCI chiplet demonstrated at OFC was co-packaged with an Intel CPU but can also be integrated with next-generation CPUs, GPUs, IPUs, and other system-on-chips (SoCs).
This first OCI implementation supports up to 4 terabits per second (Tbps) bidirectional data transfer, compatible with peripheral component interconnect express (PCIe) Gen5. The live optical link demonstration featured a transmitter (Tx) and receiver (Rx) connection between two CPU platforms over a single-mode fiber (SMF) patch cord. The CPUs generated and measured the optical Bit Error Rate (BER), showcasing strong signal quality with a 32 Gbps Tx eye diagram.
The chiplet supports 64 channels of 32 Gbps data in each direction up to 100 meters, utilizing eight fiber pairs with dense wavelength division multiplexing (DWDM). It consumes only 5 pico-Joules (pJ) per bit compared to pluggable optical transceiver modules at about 15 pJ/bit, highlighting its hyper-efficiency for data centers and HPC environments.
About Intel’s Leadership in Silicon Photonics: As a market leader in silicon photonics, Intel leverages over 25 years of internal research, pioneering integrated photonics. Intel was the first to develop and ship silicon photonics-based connectivity products with industry-leading reliability at high volumes to major cloud service providers.
Intel’s unparalleled integration using hybrid laser-on-wafer technology and direct integration yields higher reliability and lower costs. This approach enables superior performance while maintaining efficiency. Intel’s platform has shipped over 8 million PICs with over 32 million integrated on-chip lasers, boasting a laser failures-in-time (FIT) rate of less than 0.1.
These PICs have been deployed in large data center networks for 100, 200, and 400 Gbps applications. Intel is developing next-generation 200G/lane PICs to support emerging 800 Gbps and 1.6 Tbps applications, implementing a new silicon photonics fab process node with enhanced device performance and economics.
What’s Next: Intel’s current OCI chiplet is a prototype. Intel is collaborating with select customers to co-package OCI with their SoCs as an optical I/O solution.
Intel’s OCI chiplet represents a significant advancement in high-speed data transmission. As the AI infrastructure landscape evolves, Intel continues to drive innovation and shape the future of connectivity.
For more information, visit Intel Silicon Photonics.
Editor’s Note: On June 28, 2024, an incorrect reference to “state-of-the-art” was corrected in this story.
Text source: Intel
Recent Comments