Samsung Electronics has announced new memory chips that feature artificial intelligence (AI) technology, extending the company’s technological supremacy in the memory chip market.
At Hot Chips, an annual semiconductor industry conference dedicated to promoting innovative technology, the business unveiled and demonstrated how the new devices are used.
The business claims to have created HBM-PIM, a high-bandwidth memory (HBM) with AI processing capacity that would be used in data center servers.
According to Samsung, the AI processor-embedded memory chips assist central processing units (CPUs) digest data faster, as they can undertake traditionally performed data tasks.
In February, Samsung announced the industry’s first HBM-PIM chips, claiming that they are now testing the devices to improve high-speed data processing in supercomputers and other applications.
Since then, the HBM-PIM has been put through its paces in Xilinx’s Virtex UltraScale+ AI accelerator. The memory chips improved system performance 2.5 times while reducing energy usage by more than 60%, which was a good outcome.
“HBM-PIM is the industry’s first AI-tailored memory solution, which is currently being evaluated in customer AI-accelerator systems and displaying enormous commercial promise,” said Kim Nam-sung, Samsung’s senior vice president.
According to the executive, the company’s HBM-PIM may be used in a variety of applications, including supercomputers, data centers, and other AI applications that demand faster data processing.
SAP, a German technology firm, is claimed to be testing Samsung’s memory chips in its SAP HANA database management system.
When using the new memory chips, SAP’s head of HANA core research and innovation, Oliver Rebholz, predicted considerable performance gains and increased energy efficiency.
The electronics company has plans to unveil a new generation of memory chips in late 2021, the company’s first in seven years, that claims to quadruple speeds and provide the most capacity yet to keep up with the rise of data centers and artificial intelligence demands.