#What Are the Latest Developments from SK hynix in HBM4E Memory Technology?
SK hynix has accelerated its timeline for shipping 12-layer HBM4E memory samples, starting shipments earlier than expected. Initially slated for the second half of 2026, the company has now begun shipping these samples as of June 18, 2023. These advanced samples come with a significant 48GB capacity and impressive speeds of up to 16 Gbps per pin. Notably, power efficiency has improved by over 20% compared to the previous generations, marking a competitive advancement in the hardware race crucial to Artificial Intelligence (AI) applications.
The enhancements in these new memory samples can largely be attributed to SK hynix’s innovative Advanced MR-MUF technology, which effectively reduces heat resistance by 17%. Given that stacking memory layers generates substantial heat, this technology enables reliable performance under the demanding conditions of sustained AI workloads.
#How Does SK hynix Compare to Competitors in the HBM Market?
In an interesting twist, Samsung Electronics announced its own 12-layer HBM4E sample shipments just weeks prior, on May 29, making them the first to market. Samsung's samples share similar specifications, boasting bandwidth capabilities of up to 3.6 TB/s. Currently, SK hynix maintains a significant market share, capturing around 57% of industry revenue, with both companies eyeing 2027 for volume production of these innovative memory solutions.
#What Role Does Nvidia Play in This Advancements?
Nvidia plays a pivotal role in the dynamics of the HBM memory market, as SK hynix serves as a primary supplier of HBM memory for Nvidia’s AI GPUs. Both companies are actively engaged in the customer qualification phase, which will help define production allocation for upcoming years. The specific feature of a 48GB capacity per stack results in fewer stacks required for each GPU package. Alternatively, this also allows for an increase in total memory capacity across accelerators without enlarging the physical dimensions of the hardware, optimizing design and performance.