South Korean startup DeepX has announced its second-generation AI accelerator chip, the DX-M2, designed for edge devices. This chip aims to enable inference of large language models with up to 100 billion parameters while consuming less than 5W of power. The DX-M2 features significantly increased memory bandwidth and improved power efficiency, partly due to a move to a 2nm process node. DeepX is also collaborating with Samsung to integrate its chip with processor-in-memory (PIM) technology, which simulations suggest could boost performance by up to 6.7x for specific models like Llama3.1-8B. AI
IMPACT Enables more powerful AI capabilities on edge devices, potentially accelerating adoption in robotics, drones, and AI PCs.
RANK_REASON New product release from an AI chip startup with significant performance claims and industry partnerships. [lever_c_demoted from significant: ic=1 ai=0.7]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
