SK Hynix Announces World's First 48GB 16-Layer HBM3E Memory, Also Developing Next-Gen PCIe 6.0 SSD and UFS 5.0 Storage



South Korean semiconductor manufacturer SK Hynix has unveiled the industry's first 16-layer HBM3E product, with samples expected to be available as early as early 2025.

[LIVE] KEYNOTEㅣSK AI SUMMIT 2024ㅣDay 2 - YouTube


SK hynix announces the world's first 48GB 16-Hi HBM3E memory — Next-gen PCIe 6.0 SSDs and UFS 5.0 storage are also in the works | Tom's Hardware
https://www.tomshardware.com/pc-components/storage/sk-hynix-announces-the-worlds-first-48gb-16-hi-hbm3e-memory-next-gen-pcie-6-0-ssds-and-ufs-5-0-storage-are-also-in-the-works

SK Hynix has just announced that it will begin mass production of the world's first 12-layer HBM3E products in September 2024. HBM3E is a version of High Bandwidth Memory (HBM), and in addition to SK Hynix, it is also sold by Samsung and Micron .

Samsung announces ultra-high-speed memory 'HBM3E 12H' with a transfer speed of 1280GB/s and a capacity of 36GB, enabling faster AI learning and increased parallel execution of inference - GIGAZINE



SK Hynix's newly announced product is compatible with the 16-layer HBM3E standard, which boasts a capacity of 48GB per stack (3GB per die). Like the 12-layer HBM3E, it is intended for AI, and is said to have a 32% improvement in inference performance and an 18% improvement in training performance compared to the 12-layer HBM3E. SK Hynix plans to provide samples of 16-layer HBM3E products by early 2025.

SK Hynix is already working with TSMC to mass-produce HBM4, and plans to begin shipments in 2026. NVIDIA also plans to launch its HBM4-based 'Rubin' platform in the fourth quarter of 2025, so technology media Tom's Hardware points out that 'the market is moving toward HBM4, and SK Hynix's 16-layer HBM3E may be short-lived.'

NVIDIA announces AI accelerator upgrade plans, Blackwell Ultra to debut in 2025 and Rubin to debut in 2026 - GIGAZINE



In addition to memory, SK Hynix is also actively working on PCIe 6.0 SSDs, high-capacity QLC-based eSSDs for AI servers, UFS 5.0 for mobile devices, and is developing LPDDR5/6 memory to power future laptops and handhelds.



in Hardware, Posted by log1p_kr