SK Hynix Showcases AI Memory Solutions

New CXL-Based Memory and In-Memory Computing Aim to Boost Data Center Efficiency SK hynix joins 'OCP Global Summit'

SK hynix is accelerating its push into the global server market by putting forward next-generation memory solutions designed to overcome the limits of artificial intelligence infrastructure. Beyond simply supplying memory, the company’s strategy is to solidify its leadership in the AI era by demonstrating integrated solution capabilities that improve the efficiency of entire systems.

SK hynix said on Oct. 16 that it participated in the 'OCP (Open Compute Project) Global Summit,' held from Oct. 12 to 15 at the San Jose Convention Center in California, the United States, where it showcased next-generation memory technologies. At the event, which brought together IT and semiconductor companies from around the world, SK hynix devoted a substantial portion of its exhibition space to a 'system demo' section, underscoring its commitment to building a next-generation server ecosystem.

At the center of the exhibition was the second-generation 'CMM-DDR5' product based on the Compute Express Link (CXL) interface. The 256GB product features the CXL 3.2 standard and is characterized by bandwidth that is more than twice that of the previous generation. In particular, it improves energy efficiency by about 30%, making it a potential alternative for addressing the chronic power consumption problem at data centers.

CMM-DDR5 supports the use of a shared memory pool by multiple servers according to need. By enabling flexible allocation of limited memory resources, this helps maximize the efficiency of resource utilization across the entire system.

SK hynix also introduced the 'CMM-Ax' solution, which performs computing directly within memory. The technology dramatically reduces the power and cost burden on data centers by cutting down the process of data moving back and forth between the CPU and memory. Alongside this, the next-generation system 'MARS,' which integrates partner companies’ solutions into a single rack to improve efficiency, also drew attention from visitors.

Presentation sessions for technology sharing were also held at the site. Kim Ho-sik, vice president in charge of Memory System Research at SK hynix, and working-level staff explained in detail the structure of memory-centric server systems and the performance indicators of key technologies, emphasizing the company’s technology leadership.

SK hynix plans to continue technological innovation so that AI technology can keep advancing with memory at its center. The solutions unveiled at this summit are expected to establish themselves as core infrastructure for future high-performance computing and large-scale data centers.