Samsung has confirmed that it has successfully verified Compute Express Link (CXL) memory operations in a real user environment with open-source software vendor Red Hat.
The tech giant says advances in generative AI, autonomous driving and in-memory databases have all increased data throughput and memory requirements, creating demand for greater memory bandwidth and growth capacity.
CXL promises to address some of the current limitations facing data centers, including speed, latency and expandability.
Samsung verifies CXL for data centers
CXL is a unified standard that connects processors such as CPUs, GPUs and memory devices via a common PCIe interface, but with lower latency and higher bandwidth than before.
Yongcheol Bae, EVP Memory Product Planning of Samsung Electronics, said: “Our CXL collaboration with Red Hat is an exemplary example of collaboration between advanced software and hardware, which will enrich and accelerate the CXL ecosystem as a whole.”
Samsung's CXL memory in this case is optimized for Red Hat Enterprise Linux (RHEL) 9.3 and has verified memory recognition, reads, and writes in Red Hat's KVM and Podman environments.
Samsung and Red Hat are now working on producing a “RHEL 9.3 CXL Memory Enabling Guide” to help users use Samsung's CXL memory on RHEL 9.3 to build high-performance computing systems.
Marjet Andriesse, SVP and head of Red Hat Asia Pacific, said: “This is an important milestone in the integration of hardware and software to build an open-source ecosystem for next-generation memory development.”
Andriesse added that the move is important because it “opens up the applicability of the CXL Memory Expander for IaaS and PaaS-based software from Red Hat.”
The CXL Consortium now has an extensive list of founding members, board members and contributing members, but the technology is one that was primarily spearheaded by Intel. CXL 1.0 and 2.0 are based on PCIe 5.0, but in 2022 CXL 3.0 was added, which uses PCIe 6.0.
In mid-2021, Samsung unveiled what it called the “industry's first memory module to support the new CXL interconnect standard” – a module with AI and HPC workloads in data centers in mind. However, the ensuing 18 months have seen significant progress in artificial intelligence, underscoring the company's commitment to developing the hardware at the same pace.