Samsung has begun shipping samples of its HBM4E memory to major global customers, enabling Samsung to extend its high-bandwidth memory roadmap and address the increasing demand for AI infrastructure.
The 12-layer HBM4E delivers a stable pin speed of 14Gbps, scalable to 16Gbps, representing an increase of performance of 20% over HBM4. Memory bandwidth reaches up to 3.6TB/s per stack, which will maximize computing performance targeting large language models (LLMs) and next-generation AI infrastructure.
The initial 12-layer configuration ships in a 48GB capacity and even plans to add 32GB (8-layer) and 64GB (16-layer) options based on customer demand.
The HBM4E uses Samsung’s 6th-generation 10nm-class DRAM process (1c) and Samsung Foundry’s 4nm logic base die. Advanced low-power design and optimized packaging improve energy efficiency by 16% and thermal resistance by more than 14% over the prior generation.
Samsung plans to begin mass production of HBM4E later this year.
Leave a Reply