FuriosaAI Ditches GPU Playbook For 2nm Broadcom-Built Inference Chip, Claims HBM4/E Bandwidth Beats Even The Most Efficient GPUs
✦ NabkaNews BriefAuto-summarized from multiple outlets · verify with the source
FuriosaAI has reportedly switched to a 2nm inference chip built by Broadcom, claiming it offers better bandwidth than the most efficient GPUs. Other developments in the industry include the introduction of new AI chips and memory controllers, such as Rambus' HBM4E memory controller and Tensordyne's 3nm Napier AI chip. Broadcom has also made separate announcements, including the shipment of a 7nm switch and a target price hike by an analyst.
Full coverage
1234