During the annual Intel Vision conference, CEO Pat Gelsinger unveiled an ambitious roadmap that places generative artificial intelligence (genAI) at the forefront of Intel’s strategy.
Intel is focusing on its new Gaudi 3 GPU for training and running large language models (LLMs) essential for genAI in data centers. Additionally, Intel is introducing the Xeon 6 processors with neural processing units (NPUs) for workstations, PCs, and edge devices to support smaller, customized LLMs.
Intel aims to offer cost-effective chips with a user-friendly ecosystem compared to Nvidia’s offerings.
Gelsinger highlighted the superior performance of Intel’s Gaudi 3 AI accelerator over Nvidia’s H100 GPU, emphasizing better inference and power efficiency at a lower cost. The Gaudi 3 also excels in training various LLMs up to 50% faster than the H100.
According to IDC Research, the demand for server and storage infrastructure for training large LLMs will drive significant growth in the AI hardware market, reaching $41.8 billion by 2026.
GenAI is projected to fuel the adoption of on-device AI chipsets in PCs and mobile devices, with Intel’s Xeon chips and NPUs expected to power these devices. The upcoming Core Ultra processor, Lunar Lake, will offer high performance for genAI applications.
Source: www.computerworld.com