How does Samsung storage technology help transform artificial intelligence?

With the emergence of emerging large-scale artificial intelligence applications such as Midjournal, Google Bard, and ChatGPT/GPT-4, artificial intelligence (AI) and machine learning (ML) applications are disrupting thousands of industries and are expected to completely change past workflows. These artificial intelligence applications exhibit intelligence and behavior similar to humans, which cannot be separated from advanced memory technology. Samsung Semiconductor has unique technical advantages in this field: new memory, Solid-state drive, and special interface designed to break through the bottleneck of traditional hardware performance< Br/>




&Nbsp;
In order to meet the needs of artificial intelligence and machine learning applications, Samsung has launched a series of innovative technologies. Among them, PIM technology is a technology that directly integrates data processors into HBM DRAM, which can transfer some data computing work from the host processor to memory, thereby reducing data movement and improving the energy efficiency and data processing efficiency of artificial intelligence applications. PIM technology significantly improves the efficiency of artificial intelligence applications, with a performance improvement of 3.5 times.
In addition to moving a large amount of data in memory, artificial intelligence models also need to write and access disk data more frequently, but the data blocks read and written each time are smaller compared to traditional applications. In order to meet the special needs of AI applications, Samsung has introduced two innovative technologies in the traditional Solid-state drive (SSD). Memory semantic SSD has been optimized specifically for reading and writing small data blocks, greatly improving random read speed while reducing latency. SmartSSD is a computing capable hard drive that also integrates a data processor. By processing data closer to the original data store, SmartSSD can bring significant performance improvements to data intensive applications, especially for large-scale AI applications.
The Compute Express Link (CXL) interface is an open standard that enables high-capacity high-speed connections between processors and storage. Taking advantage of the CXL standard, Samsung Semiconductor has created innovative solutions to help improve the performance and energy efficiency of large-scale applications. Samsung has launched CXL storage extender technology. By utilizing this technology, the storage capacity of the server can be expanded to tens of terabytes, and the bandwidth can be increased to hundreds of gigabytes per second. In addition, Samsung has also launched CXL-PNM technology. CXL-PNM technology achieves data processing functions closer to memory, which can increase the loading speed and capacity of artificial intelligence models by two times and four times< Br/>


&Nbsp;
In short, Samsung Semiconductor is well aware that artificial intelligence applications will become increasingly important in the 21st century. Therefore, we will continue to focus on developing new generation technologies for large-scale artificial intelligence and machine learning applications, and release necessary software and simulators to support the implementation of these technologies. Because, in Samsung's view, the innovation of artificial intelligence has just begun.