Characterization of HBM Power Consumption (ETH Zurich, NYU)

Researchers model empirical power consumption across physical HBM2 DRAM chips

By GPU Data Hub Desk · GPU Data Hub·Editorial standards·14 Oct 2026, 15:27

Academics from ETH Zurich and New York University have published research evaluating the power consumption of high bandwidth memory architecture. The study provides empirical characterisation and modeling derived from tests on physical HBM2 DRAM hardware.

Original source

This summary was written by the GPU Data Hub desk from reporting published by Semiconductor Engineering on 14 Oct 2026, 15:27. Read the full article at the original publisher.

Read Original Source →

Topics

HBM2 memory power consumption semiconductor research

LinkedIn X

More from GPU Data Hub

Latest →

GlobalFoundries to build and operate two new chip fabs in NY as part of $21bn US-Japan deal

Contract chipmaker GlobalFoundries is set to construct and manage two semiconductor fabrication plants in New York state under a $21bn bilateral initiative between the United States and Japan. Commercial operations at the initial manufacturing facility are projected to commence in 2029.

Semiconductors·2h ago·Data Center Dynamics

SK hynix to place custom tiny CPU cores in its SSD controllers to ease Arm load as drive speeds rocket

SK hynix has revealed proprietary micro-cores designated as H-TPU for deployment inside its next-generation enterprise solid-state drive controllers. These specialised units are designed to handle routine input-output workloads, reducing processing strain on the primary Arm cores as drive bandwidth escalates.

Semiconductors·4h ago·Tom's Hardware

Analyzing Intel's EMIB and Foveros advanced packaging roadmaps

An industry assessment reviews the current status and future milestones for Intel's EMIB and Foveros advanced packaging lines. The semiconductor manufacturer pioneered commercial adoption of these techniques between 2017 and 2019 to interconnect heterogeneous chiplets. The roadmap outlines how Intel intends to refine these packaging technologies to maintain performance scaling in future processor generations.

Semiconductors·5h ago·Tom's Hardware

32GB of DDR5 6000 RAM now costs $450

Retail pricing for 32GB DDR5 6000 memory kits has escalated to $450 amid ongoing production and inventory bottlenecks. Available supplies of consumer-grade desktop DIMMs remain limited across the channel. The tight inventory continues to elevate acquisition expenses for builders and upgrade markets.

Semiconductors·7h ago·Tom's Hardware

ASML says it's getting order interest through 2030 as profit smashes expectations

ASML has surpassed quarterly profit forecasts and reported early customer interest extending as far out as 2030. The Dutch lithography equipment specialist noted that prospective buyers are already exploring tool allocations for 2029 and 2030. The sustained pipeline provides investors with greater visibility into structural, long-term semiconductor manufacturing demand.

Semiconductors·10h ago·CNBC Technology

Generative AI On A General-Purpose Neural Processor (GPNPU)

A novel processor concept brings together scalar, vector, and matrix execution under a single instruction set architecture tailored for generative artificial intelligence. The approach eliminates separate computing silos by consolidating varied operations into a single programming model based on standard C++. Such unified silicon aims to streamline software development while efficiently handling diverse neural network workloads.

Semiconductors·11h ago·Semiconductor Engineering