SK Hynix has taken a step ahead in the development of next-generation intelligent memory semiconductors by unveiling a memory that maximizes data processing speed by integrating memory semiconductors and computing functions into a single unit.

SK Hynix announced on the 16th that it has developed PIM (Processing-In-Memory), a next-generation memory semiconductor equipped with computing capabilities.

PIM is a next-generation technology that adds computational capabilities to memory semiconductors, enabling it to solve data congestion issues in the fields of artificial intelligence (AI) and big data processing.

Until now, the general perception has been that memory semiconductors are responsible for data storage, while non-memory semiconductors such as CPUs or GPUs handle processing functions—which are akin to the human brain. Breaking this notion, SK Hynix has consistently researched 'next-generation smart memory' capable of computation, and has now unveiled its first results.

SK Hynix plans to unveil its PIM development achievements at the '2022 ISSCC,' the world's most prestigious conference in the semiconductor field, to be held in San Francisco, USA, at the end of this month. The company expects that as this technology evolves in the future, 'memory-centric computing,' in which memory semiconductors play a central role in ICT devices such as smartphones, will also become possible.

In addition, SK Hynix has developed a sample of 'GDDR6-AiM (Accelerator in Memory)' as the first product to incorporate PIM. This product adds computational capabilities to GDDR6 memory, which processes data at a speed of 16 gigabits per second (Gbps). When this product is installed alongside a CPU or GPU instead of standard DRAM, the speed of specific computations can be up to 16 times faster. Going forward, GDDR6-AiM is expected to be utilized in machine learning, high-performance computing, and the computation and storage of big data.

In particular, this product operates at 1.25V, which is lower than the existing operating voltage of 1.35V for GDDR6. Additionally, the PIM, which performs internal computations, reduces data transfer to the CPU/GPU, thereby lowering power consumption by the CPU/GPU. As a result, energy consumption is reduced by approximately 80% compared to existing products. The company believes that this will enable it to achieve results in terms of ESG management by reducing carbon emissions from the devices in which the product is installed.

Meanwhile, SK Hynix plans to showcase technology combining GDDR6-AiM and AI semiconductors in collaboration with SAPEON, an AI semiconductor company recently spun off from SK Telecom.

Sapion CEO Ryu Su-jeong stated, “The use of artificial neural network data has increased rapidly recently, requiring computing technology optimized for these computational characteristics,” adding, “By combining the technologies of both companies, we will maximize efficiency in terms of data computation, cost, and energy usage.”