Samsung published their research on solving KV Cache scaling challenges with CXL-based memory pooling.Question: Can a CXL memory pool support large-scale KV Cache offloading while maintaining performance comparable to DRAM?Evaluation: CXL memory pooling can deliver both near-DRAM performance and substantial memory scalability for AI inference workloads. Therefore, a CXL memory-based pool can be effectively utilized as a memory expansion solution for KV cache offloading.I would highly recommend reading Samsung's paper for more info on the specific system configurations used, and various performance evaluations.