SK hynix presents rack-scale memory pooling for AI systems

SK hynix presented its Memory-Centric AI Rack System, or MARS, at the OCP Global Summit. The design separates memory from individual servers and pools it across a rack through Compute Express Link (CXL) fabrics. SK hynix says the approach is intended to address capacity and cost constraints that limit the scaling of large AI models. The supplied material does not include deployment details or quantified performance results.