MemVerge Demonstrates How Compute Express Link (CXL) Memory Expansion Will Close the Gap Between CPU and Memory Performance
MemVerge, pioneers of Big Memory software, is attending MemCon to demonstrate the capacity and performance benefits of Compute Express Link (CXL) memory. The demo features a server running an online transaction processing (OLTP) benchmark, with Memory Machine software from MemVerge that provides tiering for system DIMM and E3.S CXL memory expansion from Micron. Adding CXL memory boosted transactions per second by 15%, with the addition of Memory Machine software boosting performance another 12%.[1] The results highlight the potential to close the gap between CPU and memory performance due to memory capacity and performance per core limitations — known as the “memory wall” problem — by supplementing DIMM bandwidth and capacity with CXL expansion memory.
“CXL allows capacity and bandwidth expansion to address the memory wall. It also adds a new paradigm to Memory Tiering,” says Raj Narasimhan, senior vice president and general manager of Micron’s Compute and Networking Business Unit. “Our collaborative demo with MemVerge at this week’s MemCon event showcases that the capacity and bandwidth unlocked by CXL translates to significantly better application performance.”
Latest Insights: What Techniques Will Deliver for Measuring Attention in 2023?
Additionally, CXL memory expansion will improve workload performance while keeping costs down and creating new data center system architectures through the application of memory and storage hierarchies. These results will enable greater flexibility in data center design and use to support data-intensive workloads.
According to Charles Fan, CEO and co-founder of MemVerge, application transparency will be crucial to the success of CXL. “Our configuration running at MemCon demonstrates the ability of the CXL ecosystem to deliver tiered memory systems that can be deployed without modification to apps.”
AiThority: How Generative AI is Transforming Audio Content
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.