We collaborated with SMART Modular Technologies to showcase our Leo CXL™ Smart Memory Controller in a real-world setup that delivered up to 2TB of additional CXL memory with SMART CXL add-in cards.
In the demo, Large Language Model inferencing tasks were run using FlexGen, achieving 5.5× higher throughput and 90% GPU utilization—demonstrating the efficiency gains CXL brings to AI workloads. The demo also featured COSMOS, providing unprecedented observability for data center performance.