New Product Release | Advantech's new generation CXL 2.0 memory revolutionizes data center efficiency!

In October 2024, Advantech announced the launch of the next-generation SQRAM CXL 2.0 Type 3 memory module. Compute Express Link (CXL) 2.0 is the next evolution of memory technology, which can achieve memory expansion and acceleration through high-speed, low-latency interconnection to meet the needs of large-scale AI training and high-performance computing clusters . CXL 2.0 is built on the basis of the CXL specification and introduces advanced features such as memory sharing and expansion, making resource utilization in heterogeneous computing environments more efficient, which has attracted widespread attention in today's high-performance computing field.

Expandable memory through E3.S 2T form factor

In traditional memory architectures, fixed memory allocation often leads to low resource utilization and bottlenecks when processing large amounts of data. Now, the E3.S specification based on the EDSFF standard has brought revolutionary changes. Equipped with CXL 2.0 memory modules, these limitations can be broken through and dynamic resource management can be achieved. This not only improves performance, but also reduces costs by maximizing the use of existing resources.

High-speed interconnection through PCIe 5.0 interface

CXL memory modules operate over the PCIe 5.0 interface. This high-speed connection ensures that data transfers remain fast and efficient even as memory is scaled across different systems. With speeds up to 32 GT/s per lane, the PCIe 5.0 interface enables CXL memory modules to deliver the bandwidth required for data-intensive applications. Increased capacity will result in better performance and higher memory bandwidth without the need for more servers and capital expenditures.

Memory Pool

CXL 2.0 enables multiple hosts to access a shared memory pool, optimize resource allocation, and improve overall system efficiency. Through CXL's memory pool technology, computing components such as CPUs and accelerators of multiple servers on the same rack can share memory resources, reduce resource redundancy, and solve the problem of low memory utilization.

Hot-swappable and scalable

CXL memory modules can be added or removed from the system without shutting down the server, enabling instant memory expansion. For data centers, this means the ability to scale memory resources as needed, ensuring optimal performance without interruption.

Key Features

• Dimensions: EDSFF E3.S 2T

• CXL 2.0 is compatible with PCIe-Gen5 and runs at speeds up to 32 GT/s

• Support ECC error detection and correction

• PCB: 30μ'' gold finger

• Working environment: 0 ~ 70°C (Tc)

• Compliant with CXL™ 1.1 and CXL™ 2.0 standards

• Target applications: Data centers, machine learning, AI training, edge computing

Advantech’s SQRAM CXL 2.0 memory expansion series is currently undergoing server testing procedures. Early samples are ready, and the company plans to launch mass production in the first quarter of 2025. For more information, please call 400-001-9088.