proomt

Search

Search posts, papers, and topics

All posts

Hacker News front pageSeo Jong-Gap2 min readintro

Samsung is expected to more than double output of its HBM4 and HBM4E DRAM

Read on sedaily.comHN 543·426 comments

Summary

Samsung plans to more than double its HBM4 and HBM4E output next year, raising glass‑carrier cleaning to 50 k sheets/month and wafer input to ~250 k per month. The HBM4 family’s share of shipments is expected to jump from ~40 % to ~80 % as the higher‑layer HBM4E ramps up.

  • Samsung will increase glass carrier cleaning to 50,000 sheets/month to support HBM4/HBM4E scaling.
  • HBM4/HBM4E output is expected to more than double, with wafer input rising ~40% to 250k wafers next year.
  • HBM4's share of shipments should grow from ~40% to ~80% as HBM4E mass production ramps up.
  • HBM4 uses 10nm-class DRAM and a 4nm base die; HBM4E samples have been shipped to Nvidia.

GPU designers and data‑center engineers should track HBM supply trends because they affect high‑performance compute capacity.

4/10
architectureperformancegpu

Related reading

  2. Korea raises data breach fines to 10% of revenue

    South Korea’s privacy regulator will fine companies up to 10 % of revenue for large‑scale data breaches, a jump from the previous 3 % cap. The rule applies to intentional or grossly negligent leaks affecting 10 M+ people, with reductions for proactive security investments and rapid breach response. Companies must also notify users within 72 hours of a high‑risk exposure.

    Hacker News front pagekoreajoongangdaily.com3 minHN335113

  3. M4N VM family, now GA: Highest per-core IOPS and throughput for I/O and memory-bound workloads

    Google Cloud’s GA‑available M4N VM family pairs 5th‑gen Intel Xeon CPUs with Google’s custom Titanium offload to deliver up to 1 M IOPS, 25 GiB/s block‑storage throughput, 400 Gbps VM‑to‑VM bandwidth, and a 26 GB/vCPU memory ratio (up to 5.9 TiB RAM). The design targets memory‑bound, I/O‑intensive workloads (Oracle, SAP HANA, vector search, real‑time analytics) and claims >20 % TCO reduction for…

    Google Cloud Bloggoogle.com5 min

  5. SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops

    SiliconBench benchmarks nine Apple‑Silicon LLM serving engines on unified‑memory desktops, measuring speed, memory usage, and output fidelity across Qwen3, Qwen3.5, and Gemma‑4 models. It finds vllm‑metal doubles throughput at modest concurrency, memory budgets often fail to preserve headroom, and only three stacks satisfy all fidelity and model‑coverage requirements, with tensor‑parallel scaling…

    Hugging Face Daily Papersarxiv.org1 minpaper

  6. Custom home server built from spare parts

    A DIY home NAS built from a repurposed Zotac mini‑PC, 6×8 TB WD Red drives in a RAIDZ2 ZFS pool, encrypted boot, Wi‑Fi, custom Noctua cooling, and a front‑panel ESP32‑S3 touch screen driven by LVGL. Runs Kubuntu 24.04 with Docker‑Compose services behind Caddy, plus a self‑written dashboard and API.

    Hacker News front pageasmat.ca7 minHN8949