SK hynix begins mass production of next-gen AI server memory designed for Nvidia's Vera Rubin platform

SK hynix Inc. said Monday it has begun mass production of a next-generation memory module designed for artificial intelligence servers, as it seeks to strengthen its position in the AI infrastructure market.

The company said the 192GB SOCAMM2 module is based on its sixth-generation 10-nanometer-class LPDDR5X low-power DRAM technology. The module was designed particularly for use with Nvidia Corp.'s Vera Rubin AI platform.

The module adapts mobile-oriented low-power memory for server environments and is designed to serve as a primary memory solution for next-generation AI servers.

According to SK hynix, the product delivers more than double the bandwidth and over 75 percent improved power efficiency compared with conventional RDIMMs, or registered dual in-line memory modules, making it suitable for high-performance AI operations.

The South Korean chip giant said the new product is expected to help resolve memory bottlenecks in the training and inference of large language models with hundreds of billions of parameters, significantly improving overall system performance.

"By supplying the 192GB SOCAMM2, SK hynix has established a new standard for AI memory performance," said Kim Joo-sun, president and head of AI infrastructure at the company. (Yonhap)