Semiconductor News & Analysis Feed
31 articles
2026-05-27
letsdatascience.com
2026-05-27
Let's Data Science
INFRASTRUCTURE
broadcom
furiosaai
ai accelerators
advanced packaging
Broadcom Adds FuriosaAI for Custom AI Accelerators
2 sources
|
May 27, 2026
6.8
Relevance Score
Photo: image.theregister.com · rights & takedowns
QUICK SUMMARY
Hide
Broadcom has added South Korea's FuriosaAI to its roster of partners building AI accelerators on Broadcom's packaging and networking technology, The Register reports
2026-05-27
www.indexbox.io
2026-05-27
IndexBox
Thermal management has become the primary limiting factor in high-performance computing and AI accelerator packaging. Contemporary packages combine high-power ASICs with multiple High Bandwidth Memory stacks on a silicon interposer, resulting in closely linked thermal and mechanical interactions. Thermal crosstalk between dies increases HBM junction temperatures, while mismatched coefficients of t
2026-05-26
www.chosun.com
2026-05-26
조선일보
By Seongmin K
Published 2026.05.27. 00:34
Concept diagram of SK Hynix's iHBM technology. ICE, a thermal control element, serves as a passage for heat dissipation. /SK Hynix
SK Hynix unveiled on the 26th its "iHBM" technology, which significantly reduces heat generation in high-bandwidth memory (HBM) used in AI accelerators.
As demand for AI computations rises, the workload of HBM mounted on AI
2026-05-26
www.tomshardware.com
2026-05-26
Tom's Hardware
__fail__
2026-05-19
www.openpr.com
2026-05-19
openPR.com
__fail__
2026-05-19
dataconomy.com
2026-05-19
Dataconomy
__fail__
2026-05-12
www.openpr.com
2026-05-12
openPR.com
__empty__
2026-05-12
semiengineering.com
2026-05-12
Laura Peters
Multi-die assemblies greatly increase the number of things that can go wrong, and the d...
2026-05-11
www.techradar.com
2026-05-11
TechRadar
Copy linkFacebookXWhatsappRedditPinterestFlipboardThreadsEmailShare this article0Join the conversationFollow usAdd us as a preferred source on GoogleNewsletterSubscribe to our newsletterSkymizer claims giant AI models no longer need hyperscale GPU infrastructureOld 28nm chips suddenly power massive language models at surprisingly low wattageThe HTX301 squeezes 384 GB of memory into a single PCIe a
2026-05-07
wccftech.com
2026-05-07
Wccftech
__fail__
2026-05-06
www.openpr.com
2026-05-06
openPR.com
__fail__