Visit our kiosk (Booth no. 725A) to view CXL video demos and explore how CXL can expand your memory footprint.
Astera Labs and SMART Modular Technologies: Boost AI Inferencing with CXL
In this demo, Astera Labs and SMART Modular Technologies are showcasing how showcasing how CXL memory expands memory capacity and improves I/O efficiency, significantly boosting AI inferencing performance. Using Leo CXL Smart Memory Controllers and SMART Modular CXL Cards, we’re able to deliver 5.5x higher throughput and 90% GPU utilization.
Cadence – Cadence IP for CXL3.X
Cadence demonstrated its design IP and VIP for CXL at CXL DevCon, 25’
Montage Technology – CXL Memory Expansion: Real-World Performance Demo Using Montage’s CXL MXC Controller
In this demo, Montage compares a traditional 1TB DRAM-only server to a hybrid 1TB system with 512GB DDR5 DRAM and 512GB CXL memory provided by four RDIMMs through Montage’s MXC add-in cards. Configured in flat memory mode, the system runs MongoDB and YCSB across sockets to simulate real-world workloads. Despite using CXL-attached memory, the system achieves 95–100% of the throughput delivered by DRAM alone, with only 5–10µs added application-level latency. Montage’s CXL MXC controller contributes just 70ns of delay and provides a scalable, production-ready, and cost-efficient path to expand memory for data-intensive environments.
Panmnesia – Running RAG on Panmnesia’s CXL-Enabled AI Cluster Including CXL 3.x Switches
Panmnesia is a company developing link solutions for AI infrastructure. This demo video showcases Retrieval Augmented Generation (RAG) running on Panmnesia’s CXL-enabled AI cluster, which includes CXL 3.x switches. In the demonstration, Panmnesia accelerates RAG by running the large language model (LLM) on CXL-enabled GPU nodes, while performing vector search on CXL-enabled memory nodes.
For more information about Panmnesia’s CXL technologies, please visit the official website: https://panmnesia.com/.
Rambus – CXL 3.0 Controller IP with XpressAGENT Demo
Discover how the Rambus CXL 3.0 Controller IP, combined with the XpressAGENT embedded protocol analyzer, enables real-time inspection, validation, and debugging of CXL.mem transactions in modern data center environments. This demo showcases the setup using an Intel Granite Rapids-based server and a Rambus Gen5 endpoint device, walks through key steps such as link identification, packet analysis, and live traffic monitoring. See how the Rambus CXL 3.0 Controller and XpressAGENT together offer deep visibility into the CXL protocol to help streamline and accelerate your CXL memory expansion projects.
SMART Modular – SMART Modular Technologies Overview on CXL
In this video, SMART Modular’s senior product manager for advanced memory, Torry Steed, highlights SMART Modular’s CXL modules and add-in cards utilizing the CXL 2.0 standard as well as showing two partner demonstrating the add-in cards in servers.
Learn more by visiting https://www.smartm.com/product/list/cxl-memory.
UnifabriX – Revolutionizing Data Center Efficiency: UnifabriX Memory over Fabrics in Action
Join Oren Benisty, VP of Marketing and Business Development at UnifabriX, for an exclusive deep dive into the future of Memory over Fabrics architecture. Discover how UnifabriX is transforming data center performance with its groundbreaking CXL-based Memory Pooling and Memory Sharing solution.
This presentation will unveil:
- Key technology highlights that power UnifabriX’s innovation
- The TCO model that underpins scalable, efficient memory disaggregation
- Real-world use cases demonstrating dramatic improvements in resource utilization, cost-efficiency, and workload flexibility
Whether you’re a cloud architect, data center strategist, or tech enthusiast, this presentation will reshape how you think about memory in modern computing.
Wolley – Memory Expansion For AI Workstations
Modern large-language models (LLMs) are shifting focus to support larger context windows such as 128 K tokens, further driving memory-bandwidth demand to several hundreds of GB/s, beyond the 100 to 200 GB/s ceiling of today’s CPU and NPU platforms. GPUs, while popular for AI acceleration, are limited by on-package memory and rising costs, making them an increasingly expensive way to meet these requirements.
Wolley will showcase an E1.S CXL Memory Module that breaks the memory barrier plagued in the LLM applications. This CXL memory module delivers 32 GB/s to 64 GB/s bandwidth over CXL protocol by expanding the effective memory capacity through DRAM and SSD virtualization. The result is a single plug-in card that unifies memory and storage, eliminates PCIe congestion, and reduces total cost of ownership, empowering on-prem systems to run long-context LLM inference smoothly and at scale.
XConn Technologies – Demo of dynamic memory pool using CXL VCS
XConn CXL 2.0 switch supports VCS, enabling multiple hosts to access CXL memory pool.