Samsung announces it has manufactured a sample of 16-stack HBM for AI GPUs

Samsung has announced that it has manufactured a sample of its brand-new 16-stack HBM. The company manufactured the chip with hybrid bonding last week.

2

Samsung vice president Kim Dae-woo said the South Korean giant made its new 16-layer HBM with HBM3 but plans to use HBM4 to improve productivity. Due to alignment issues, Samsung was expected to use hybrid bonding for just one or two stacks of the HBM chip but applied the hybrid bonding technique to all 16 stacks.

The 16-stack HBM memory sample was made using equipment from Samsung's fab equipment subsidiary, Semes, with Kim noting that Samsung considered using hybrid bonding or thermal compression non-conductive film for HBM4, which Samsung will be sampling in 2025 and mass producing next-gen HBM4 memory in 2026.

• Read more: Samsung forms dedicated 'HBM team' to boost AI memory chip production
• Read more: Samsung to use MR-MUF technology, like SK hynix, for its new HBM products
• Read more: Samsung to unveil world's first HBM3E 12-High memory at NVIDIA GTC 2024
• Read more: NVIDIA is qualifying Samsung's new HBM3E chips, will use them for B200 AI GPUs
• Read more: HBM supply growth estimated at 260% in 2024, 14% of the DRAM industry
• Read more: NVIDIA Blackwell B200 AI GPU: multi-chip GPU die, 208B transistors, 192GB HBM3E

Samsung's just-announced HBM3E memory uses TC-NCF technology, while hybrid bonding can be better as they can add more stacks compactly without needing through-silicon-via (TSV), which uses filler bumps to connect. The core die DRAM on the HBM memory chips can also become thicker with the same technique, notes TheElec.