SK hynix unveils new tech for 'dream memory chip' to store data, perform calculations for AI

SK hynix unveiled a new technology that will be used to create a "dream memory chip" that is capable of storing data and performing calculations for AI.

2

The South Korean memory giant unveiled its new technology during the International Memory Workshop (IMW 2024) held from May 12-15 at the Walkerhill Hotel in the Gwangjin district of Seoul, South Korea. The new technology enhances the accuracy of Multiply Accumulate (MAC) operations in Analog Computing in Memory (A-CIM) semiconductors using oxygen diffusion barrier technology.

MAC operations are critical for the high-speed multiplication and accumulation processes required in artificial intelligence (AI) inference and learning. SK hynix's recent development is a significant step for the company in the competitive field of creating a "dream memory semiconductor" that can both store information and perform calculations, passing the traditional limitations of memory-only semiconductors.

Business Korea reported the news, adding that the semiconductor manufacturing process involves etching away unwanted materials, depositing necessary substances to build uniform layers, during which performance degradation can occur due to chemical interactions. An oxygen diffusion barrier is required to prevent degradation, stabilizing semiconductor operations.

Kwon Young-jae, a team leader at SK hynix, said at the IWC 2024 event: "By preventing oxygen diffusion, we can increase the accuracy of MAC operations, which is crucial for creating CIM semiconductors. Enhancing MAC operation accuracy with oxygen diffusion barrier technology means taking a step closer to transforming memory semiconductors into devices capable of performing calculations".

Kwon also noted: "They are, but they tend to not announce their developments until their technologies have matured. What SK hynix showcased today is just a part of our overall technology efforts".

Memory semiconductors are perfect for data storage, but they simply can't perform calculations on their own... but by creating a calculating CIM is a "long-standing dream" and aspiration in the memory semiconductor industry. The development of CIM semiconductors would remove the need for data transfer from memory to the CPU for processing, which is a gigantic savings in energy... allowing for higher speeds.

This could be a game-changer for the industry, transforming memory semiconductor companies into foundry operations -- including Samsung and TSMC -- taking orders from companies like Apple, AMD, NVIDIA, Microsoft, Tesla, Qualcomm, and more.

What abhout CIM semiconductors for AI GPUs and AI accelerators? Kwon noted: "It could be possible in the very distant future. Although it's not easy, memory companies are staking their futures on developing CIM semiconductor technology and related advancements".