Achieves data processing speeds of up to 16Gbps per pin, increasing energy efficiency by over 20%

SK Hynix has supplied 12-layer samples of its next-generation high-performance AI DRAM, 'HBM4E,' to major customers. This product enhances data processing capabilities required for AI training and inference by improving both performance and power efficiency compared to the previous generation, HBM4.

SK Hynix announced on the 18th that it has supplied HBM4E 12-layer samples to major customers.

SK Hynix explained, “Based on the HBM advanced development capabilities and production know-how accumulated so far, we have showcased HBM4E 12-layer samples to our customers,” adding, “We will work closely with key customers to ensure timely mass production.”

The company stated that HBM4E achieves a data processing speed of up to 16Gbps per pin and increases energy efficiency by more than 20%. In addition, the latest interfaces and design optimizations have been applied to reduce data transmission delays and ensure stable operation even in high-bandwidth environments.

The company expects that this will enable it to increase the processing efficiency of next-generation AI data centers and large-scale computing systems.

SK Hynix applied the Advanced MR-MUF process to HBM4E to achieve a capacity of 48GB based on 12-layer stacking while simultaneously enhancing structural stability.

In particular, thermal resistance was reduced by about 17% compared to HBM4, ensuring that the memory operates stably even in high-performance computing environments.

MR-MUF is a process in which chips are stacked, and then a liquid protective material is injected and solidified to protect the circuits between the chips.

SK Hynix announced that, based on its mass production and supply experience spanning HBM3, HBM3E, and HBM4, it has been providing memory solutions tailored to customer needs, and plans to support the resolution of AI system bottlenecks and the implementation of next-generation infrastructure with HBM4E as well.

"We have laid the foundation to continuously support AI innovation by extending our accumulated technological competitiveness and mass production capabilities to HBM4E," said Ahn Hyun, Chief Development Officer (CDO). "Based on cooperation with partners, we will realize the value demanded by the market and solidify our technological leadership as a full-stack AI memory company."