Back to feed
News Story
APriority83
DeepTech深科技
1 sources

SK Hynix Bets on Next-Gen CPO: Optical Interconnects Enter Memory Interface

SK Hynix, along with several universities, published a review paper in Nature Electronics proposing to extend co-packaged optics (CPO) to the memory interface to address bandwidth bottlenecks in AI computing. The technology could reshape memory resource organization, enabling multiple AI accelerators to share a large memory pool and achieve over 100 Tb/s bandwidth with sub-1 pJ/bit energy efficiency.

SynthePulse Insight · AI deep readingMembers

SK Hynix Bets on Extending Co-Packaged Optics to Memory Interfaces: How AI Systems Organize Memory Could Be Reshaped

Version 1 · 1 source

SK Hynix, together with several universities, published a review in Nature Electronics proposing a 'photonics-centric' architecture that pushes optical interconnects to memory interfaces, targeting over 100 Tb/s bandwidth per node and below 1 pJ/bit energy consumption. This is not just a technology upgrade; it could change how AI systems organize memory and drive competition in upstream indium phosphide materials.

  • SK Hynix and other institutions published a CPO review in Nature Electronics, proposing a 'photonics-centric' architecture that extends optical interconnects to memory interfaces.
  • The long-term goals of the architecture include over 100 Tb/s bandwidth per node, below 1 pJ/bit energy consumption, and below 10 nanoseconds inter-chip latency.
  • The expansion of AI clusters has widened the gap between compute growth and interconnect bandwidth, creating a new 'bandwidth wall,' with CPO emerging as a solution.
Open section navigationCPO Review Proposes 'Photonics-Centric' Architecture

CPO Review Proposes 'Photonics-Centric' Architecture

On August 19, researchers from SK Hynix, the University of Virginia, MIT, Nanyang Technological University, and other institutions published a review paper in Nature Electronics titled 'Co-packaged optics for high-performance computing and AI,' systematically outlining the development path of CPO in high-performance computing and AI, and proposing a longer-term roadmap: pushing optical interconnects to memory interfaces.

In traditional approaches, data must first travel as electrical signals from the chip to an optical module at the edge of the device before conversion; CPO places optical components directly near the chip, shortening high-speed electrical signal transmission distances, reducing power consumption, and increasing bandwidth. SK Hynix's proposed next step is to continue pushing inward, extending optical links between AI accelerators and memory, forming a 'photonics-centric' system architecture.

In this architecture, compute chips like GPUs can connect to larger memory pools through photonic interposers, allowing multiple AI accelerators to share memory resources, alleviating packaging space and electrical interface constraints, and improving memory resource utilization. The paper's long-term goals include over 100 Tb/s bandwidth per node, below 1 pJ/bit energy consumption, and below 10 nanoseconds inter-chip latency.

Free for now

Read the full analysis

3 more sections of analysis, plus the full takeaway

Loading

Credibility boundary

This article is based on DeepTech's report on the review paper published in Nature Electronics by SK Hynix and other institutions. The paper's content, NVIDIA product deployment, HBF standard, export controls, and AXT impact are all from that report and are source claims. Specific technical metrics (such as 100 Tb/s, 1 pJ/bit) are long-term goals proposed in the paper, not achieved data.

Primary report

DeepTech深科技

Primary source