Subscribe to EDOM TECH Newsletter
2023/02/23
CXL Memory Expansion and Pooling for Cloud Servers
CXL Memory Expansion and Pooling for Cloud Servers
Our personal and professional lives create and consume large amounts of information, while compute-intensive AI and ML workloads have been hampered by memory bottlenecks. In the traditional structure, memory is directly connected to a CPU or GPU and cannot be shared by other CPUs or GPUs. Despite DRAM accounts for up to 50% of server costs, up to 25% of memory is stranded at any given time. Cloud service providers need to develop new architectures to store, analyze and deliver the growing volume of data.
Compute Express Link (CXL) is an open standard for processors, accelerators, and memory expansion. CXL memory controllers enable memory expansion, pooling and sharing solutions to reduce total cost of ownership (TCO) by increasing memory utilization, flexibility and availability. CXL memory controllers are built for data center to unlock application-specific workloads in AI training and inference, machine learning, in-memory databases, memory tiering, and multi-tenant use-cases.
Compute Express Link (CXL) is an open standard for processors, accelerators, and memory expansion. CXL memory controllers enable memory expansion, pooling and sharing solutions to reduce total cost of ownership (TCO) by increasing memory utilization, flexibility and availability. CXL memory controllers are built for data center to unlock application-specific workloads in AI training and inference, machine learning, in-memory databases, memory tiering, and multi-tenant use-cases.
Astera Labs - Leo CXL Memory Connectivity Platform
Astera Labs Leo Memory Connectivity Platform is a solution supporting memory expansion, pooling, and sharing to eliminate memory bottlenecks and decrease TCO. Leo Controllers have features including server-grade RAS features, end-to-end security, fleet management, low-latency DDRx, and seamless interoperability with all major CPU, GPU, and memory vendors.
