NuAxiSemiconductor trade intelligence

Samsung Electronics Puts Speed, Scalability and Persistence at Center of Memory Strategy

What happened

Samsung Electronics unveiled its next-generation memory strategy at the Open Compute Project (OCP) Global Summit, an IT infrastructure event held in California on Oct. 12. Shim Ho-joon, vice president of the AE Team in Samsung's DS Division Memory Business, outlined three strategic priorities: speed, scalability and persistence. The stated reason in the source is that as AI models grow, the amount of data they must process is increasing, and Shim said memory capacity expansion and data storage technologies are becoming as important as memory bandwidth. Samsung plans to improve data processing speeds and computational efficiency through eighth-generation high-bandwidth memory (HBM5), custom HBM, zHBM and processing-in-memory (PIM) technologies; to expand AI system memory capacity and improve data management efficiency using CMM-D and Compute Express Link (CXL)-based technologies; and to increase AI data storage capacity and improve management efficiency with next-generation solid-state drives (SSDs), quad-level cell (QLC) technology and flexible data placement (FDP). These are stated plans, a strategy announcement rather than achieved results. Samsung also showcased the results of its collaboration with Nvidia, displaying a Groq wafer — Nvidia's language processing unit (LPU) — and the Groq 3 LPX system; its foundry business manufactures Groq chips using its 4-nanometer process. Samsung also jointly demonstrated a CXL-based memory expansion solution with AMD, and the two companies are collaborating with global customers on designs from the early stages of product development and expanding technology validation.