SK hynix has made an announcement regarding the successful development of the latest generation of High Bandwidth Memory (HBM) technology, known as HBM3E.
This cutting-edge advancement in DRAM technology is tailor-made for AI applications, delivering unparalleled data processing capabilities. Currently, the company is in the process of evaluating samples of this revolutionary memory with its customers.
HBM represents a high-performance, high-value memory solution that takes advantage of vertical interconnections among multiple DRAM chips. This architectural innovation translates into a remarkable boost in data processing speed compared to previous DRAM iterations.
HBM3E, an extension of the HBM3 standard and the fifth iteration in this series, follows the footsteps of its predecessors HBM, HBM2, HBM2E, and HBM3.
SK hynix emphasizes that the successful development of HBM3E, which boasts world-class specifications, builds upon its role as the sole major provider of HBM3 technology in the industry.
Drawing from its extensive experience in supplying the largest volume of HBM products and its readiness for mass production, SK hynix has set its sights on commencing mass production of HBM3E in the first half of the upcoming year. This strategic move is poised to cement the company's undisputed leadership in the AI memory market.
The company takes pride in affirming that their latest offering not only meets the highest industry benchmarks in terms of speed—the paramount specification for AI memory products—but also excels in other critical categories, including capacity, heat dissipation, and user-friendliness.
In the speed department, HBM3E stands out by processing data at a remarkable rate of up to 1.15 terabytes per second. To put this into perspective, this data processing capability is equivalent to handling over 230 Full-HD movies, each with a size of 5GB, within a single second.
Moreover, HBM3E introduces a noteworthy 10% enhancement in heat dissipation through the incorporation of cutting-edge technology known as Advanced Mass Reflow Molded Underfill (MR-MUF2). This innovative addition ensures that the memory module maintains optimal performance even under demanding conditions.
Importantly, the product maintains backward compatibility, allowing it to seamlessly integrate into systems designed for HBM3 without necessitating any modifications to their existing designs or structures.
“We have a long history of working with SK hynix on High Bandwidth Memory for leading edge accelerated computing solutions,” said Ian Buck, Vice President of Hyperscale and HPC Computing at NVIDIA. “We look forward to continuing our collaboration with HBM3E to deliver the next generation of AI computing.”
Sungsoo Ryu, Head of DRAM Product Planning at SK hynix, said that the company, through the development of HBM3E, has strengthened its market leadership by further enhancing the completeness of HBM product lineup, which is in the spotlight amid the development of AI technology. “By increasing the supply share of the high-value HBM products, SK hynix will also seek a fast business turnaround.”
