SK hynix is introduced One of the industry’s first Computational Memory Solutions (CMS), it features a 512GB memory expansion module with computing capabilities and a PCIe 4.0 x4 interface with CXL protocol.
Aimed at big data analytics applications, the device is said to deliver several times better performance compared to CPUs for certain calculations.
SK Hynix’s CMS is a system-on-chip (or set of chips) with four 128GB conventional DDR memory modules and capable of accelerating machine learning KNN (k-nearest neighbor) classification algorithms, data filtering, and load balancing. ). It has a PCIe 4.0 x4 interface that supports both CXL.mem and CXL.io protocols. The CMS card also has two network connectors for connecting to other nodes.
I’m not sure about the power consumption of the card, but it does have an 8-pin auxiliary PCIe power connector that can deliver up to 150W of power.
SK Hynix said the prototype CMS is a big data analytics platform co-designed with SK Telecom that can perform data filtering and real-time data processing operations faster than various many-core CPUs.
SK Hynix did not quantify the benefits of its compute memory solution over traditional processors in its press release. However, one of the pictures shown by the company seems to show graphs and numbers that show the advantages offered by the device.
Park Kyung, Head of Memory Systems Research at SK Hynix, said: “Given that this is just a prototype, we believe we can further improve performance, and we are considering applying this technology to other applications such as big data.”
The CXL CMS prototype designed by SK Hynix was first demonstrated at the Open Compute Project (OCP) Global Summit. The company also demonstrated an optimized big data analytics platform together with SK Telecom.