Computing with CXL Memory Devices - Memory Expansion and Computational Offloading

Learn about Compute Express Link (CXL) memory devices and their transformative role in datacenter applications through this 17-minute technical presentation. Explore how CXL devices serve dual purposes in memory expansion and bandwidth enhancement when socket limitations arise. Discover the cost-saving benefits of memory pooling devices in reducing DRAM requirements, while understanding the potential for offloading compute tasks to these devices. Delve into advanced concepts including data analytics offloading, telemetry functions like hotness tracking, and the integration of compute capabilities into CXL memory devices. Follow along as Mats Öberg from Marvell explains cloud data center memory challenges, resource sharing with CXL, memory expansion tiers, hot/cold page detection, and computational memory use cases that parallel computational storage approaches.