Memory and memory utilization are increasingly becoming a critical resource for future servers. At the same time, memory media is not scaling fast enough to keep up and driving more complexity, power, and cost into system designs. In this presentation, we will describe how a new technology (Compute Express Link) will help us address them by using memory expansion solutions. We will discuss the Proof of Concept CXL based memory expander module we have designed, some early results for community feedback, and then share some single socket server hardware design concepts based on the module we will be contributing to OCP in future, and also discuss our collaboration with ecosystem partners to develop a CXL memory expansion solution. We will briefly explore the new features of CXL 2.0 and share how CXL technology is keeping pace to meet the performance and latency demands of emerging workloads for Artificial Intelligence and Machine Learning
