SK hynix launches CXL optimization for Linux memory performance
- SK hynix's HMSDK is now available on Linux, optimizing performance for heterogeneous memory systems.
- The integration increases memory bandwidth by 30% and performance by over 12%, enhancing AI memory capabilities.
- This move positions SK hynix as a leader in the AI memory market, anticipating the commercialization of CXL technology in late 2024.
SK hynix has announced the integration of its Heterogeneous Memory Software Development Kit (HMSDK) with Linux, enhancing memory performance for heterogeneous memory systems, particularly Compute Express Link (CXL) technology. This integration, revealed on September 22, 2024, aims to optimize memory control, increasing bandwidth by 30% and improving performance by over 12% without requiring changes to existing applications. The company is positioning itself as a leader in AI memory solutions, particularly as the semiconductor industry anticipates the full commercialization of CXL technology in late 2024. The HMSDK allows for effective memory management by selectively allocating memory based on bandwidth and access frequency, which relocates frequently accessed data to faster memory. This capability is crucial for the development of advanced AI systems, including large language models, which require significant improvements in both semiconductor technology and system applications. As SK hynix prepares for the release of server CPUs featuring the CXL 2.0 specification, it is currently validating CXL 2.0 memory in capacities of 96 GB and 128 GB with its customers. The company plans to begin mass production by the end of the year, which will further solidify its competitive edge in the memory market. Youngpyo Joo, Head of Software Solution at SK hynix, emphasized the importance of this Linux integration for the advancement of AI memory solutions. The company aims to foster a collaborative ecosystem that enhances software competitiveness and supports the growing demand for high-performance memory technologies in various applications.