SK hynix unveils 'iHBM' thermal solution to boost AI performance
The solution cuts thermal resistance by 30% and uses existing packaging technology to help HBM chips run more stably in AI data centers.
- On May 25, SK Hynix announced the launch of iHBM, a cooling solution embedding integrated cooling elements within high-bandwidth memory packages to reduce thermal resistance by 30% in artificial intelligence computing environments.
- Heat management has emerged as a critical bottleneck as HBM technology stacks more layers and operates at higher speeds to meet surging AI data processing demand, with peak thermal challenges at the die-to-die physical layer .
- Unlike conventional HBM designs routing heat indirectly through core dies, iHBM uses electrically non-conductive, thermally conductive silicon material to create a dedicated 'heat dissipation path' directly in the D2D PHY area where concentration is highest.
- Built on SK Hynix's market-proven Wafer Level Packaging process using Mass Reflow Molded Underfill technology, iHBM enables simultaneous packaging and testing, positioning the company for rapid high-volume production scaling.
- Deployment begins with HBM5 and subsequent products targeting high-performance computing and AI data centers, with high design compatibility allowing customers to adopt the technology with minimal system redesign, strengthening SK Hynix's AI memory leadership.
51 Articles
51 Articles
SK Hynix Unveils New Technology to Reduce HBM Heat; Creates Dedicated Heat Emission Paths to Lower Thermal Resistance by Over 30%, Enhancing AI Efficiency. On the 26th, SK Hynix unveiled iHBM technology, which dramatically reduces heat issues in High Bandwidth Memory (HBM) used in artificial intelligence (AI) accelerators. The demand for AI computing...
digitimes.com
technewstube.comSK Hynix moves cooling directly to D2D PHY to tackle rising HBM heat challenges
Heat management has become a critical challenge as HBM technology advances, with higher stacking and faster speeds to meet surging global demand for AI data processing. Efficient control of power density in the die-to-die physical layer between HBM and GPUs is now central to next-generation HBM competitiveness and data center reliability.
InsideNoVA.com
SK hynix unveils 'iHBM' thermal solution to boost AI performance
Enhances heat dissipation by integrating ICEs into the HBM packageReduces thermal resistance by 30%, ensuring stable chip operation in demanding high-temperature and high-pressure environmentsLowers barriers to adoption by leveraging market-proven MR–MUF technology, featuring high design compatibility
Similar to DDR5, where costs and power consumption are increasing in the server components, high bandwidth memory (HBM) requirements are also increasing in these areas.
Coverage Details
Bias Distribution
- 61% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
