SK hynix has announced the development of its first product based on PIM technology, which speeds up AI processing, and also introduced the first model, GDDR6-AiM.
SK hynix develops PIM, next-generation AI accelerator: demonstrates GDDR6-AiM DRAM model
Press release: SK hynix (or “Company”, www.skhynix.com) announced on February 16 that it had developed PIM *, a next-generation memory chip with computing power.
* Memory Processing (PIM): A next-generation technology that provides data overload for AI and big data by adding computing functions to semiconductor memory
It is generally accepted that memory chips store data, and a processor or graphics processor, like the human brain, processes data. SK hynix, challenging such a concept and trying to innovate in the next generation’s smart memory, has found a groundbreaking solution with the development of the latest technology.
SK hynix plans to showcase its PIM development at the world’s most prestigious semiconductor conference, the 2022 ISSCC *, in San Francisco later this month. The company looks forward to continuing efforts to innovate this technology to bring memory-focused computing, in which semiconductor memory plays a central role, one step closer to reality in devices such as smartphones.
* ISSCC: International Conference on Solid State Chains will be held virtually from February 20 to February 24 this year on “Intelligent Silicon for a Sustainable World”
For the first product that uses PIM technology, SK hynix has developed a model of GDDR6-AiM (memory accelerator *). GDDR6-AiM adds computing features to GDDR6 * memory chips that process data at 16 Gbps. Combining GDDR6-AiM with a CPU or GPU instead of regular DRAM makes a certain computing speed 16 times faster. GDDR6-AiM is expected to be adopted for machine learning, high-performance computing, and big data computing and storage.
* Accelerator: special equipment created using processing chips and computing
* Graphics DDR (GDDR): a standard graphic DRAM specification defined by the Joint Electronic Devices Engineering Council (JEDEC) and specialized for faster graphics processing, the generation of which has moved from GDDR3, GDDR5 and GDDR5X to GDDR6. It is now one of the most popular memory chips for AI and big data applications.
GDDR6-AiM operates from a voltage of 1.25 V, which is lower than the operating voltage of the existing product at 1.35 V. In addition, PIM reduces the movement of data to the CPU and GPU, reducing power consumption by 80%. This, in turn, helps SK hynix fulfill its ESG management commitments by reducing the carbon emissions of devices that use this product.
SK hynix also plans to introduce a technology that combines GDDR6-AiM with AI chips in collaboration with SAPEON Inc., an AI chipmaking company that recently split from SK Telecom.
“The use of artificial neural data has increased rapidly in recent years, requiring computing technology optimized for computational performance,” said Ryu Su Zhong, CEO of SAPEON Inc. “We strive to maximize data efficiency, cost and energy efficiency by combining technology from the two companies.”
An Hyun, head of decision development, who led the development of the latest technologies and products, said that “SK hynix will build a new ecosystem of memory solutions using GDDR6-AiM, which has its own computing function.” He added that “the company will continue to develop its business model and direction of technology development.”