Intel Introduces Next-Gen AI Chips to Boost Performance at Vision 2024 Event

Intel recently introduced its latest innovation in artificial intelligence technology: the Gaudi 3 chips, designed specifically for enhanced performance in training and running large language models. This unveiling took place at Intel Vision 2024 in Phoenix, Arizona. The Gaudi 3 chips aim to transform cloud-based infrastructure, optimizing both training and inferencing processes for large language models.

Claiming remarkable advancements over previous technologies, Intel asserts that the Gaudi 3 is 50% faster for training and 30% quicker for processing large language models compared to Nvidia’s H100 GPUs. “Innovation is advancing at an unprecedented pace, all enabled by silicon – and every company is quickly becoming an AI company,” stated Intel CEO Pat Gelsinger. He emphasized the company’s mission to integrate AI across all sectors, from personal computing to expansive data centers and edge computing. The Gaudi, Xeon, and Core Ultra platforms collectively offer a versatile suite of solutions tailored to meet the evolving demands of businesses and capitalize on the expansive opportunities that lie ahead.

The Gaudi 3 chips feature significantly increased memory and networking bandwidth, delivering four times the AI compute power compared to their predecessors, the Gaudi 2 models. Equipped with tens of thousands of accelerators, these chips support interconnection via Ethernet, enhancing collaboration and efficiency across networks. Intel positions the Gaudi 3 as a cost-effective solution suitable for enterprises aiming to experiment or scale their existing AI deployments. Customers can easily transition operations from a single node to large clusters, supporting extensive AI workloads.

“In the ever-evolving landscape of the AI market, a significant gap persists in the current offerings. Feedback from our customers and the broader market underscores a desire for increased choice,” noted Justin Hotard, Intel's executive vice president and general manager of the Data Center and AI Group. He added that enterprises must consider various factors—such as availability, scalability, performance, cost, and energy efficiency—as they navigate their AI implementations. Intel’s Gaudi 3 stands out as a competitive alternative in the generative AI space, delivering a compelling mix of price performance, system scalability, and timely value.

Among the early adopters of the Gaudi 3 technology is South Korean internet provider Naver. At the Vision 2024 event, Naver announced its commitment to utilizing Gaudi hardware to power its expansive global AI services through new large language models. The Gaudi 3 chips are set to hit the market in the second half of 2024, targeting hardware manufacturers like Dell Technologies, HPE, and Lenovo, with a wider release scheduled for later this year. Additionally, a PCIe add-in card for Gaudi 3 will enable users to connect high-speed input-output components, such as Wi-Fi cards or SSDs, launching in Q3 2024.

In conjunction with Gaudi 3, Intel provided insights into its new Xeon processors. The Xeon 6 series is designed to optimize generative AI workloads, supporting both cloud and edge applications. Intel plans to debut two Xeon 6 product lines—Sierra Forest and Granite Rapids—later this year, marking a significant upgrade from the 5th generation models that were introduced earlier.

Moreover, Intel unveiled Tiber, a comprehensive suite of solutions aimed at simplifying generative AI software deployment for businesses. Tiber is designed to equip enterprise customers with the tools necessary for securely implementing AI services, whether in the cloud or at the edge, with an anticipated release in the coming months. Intel continues to pave the way for AI technology, positioning itself as a leader in this rapidly evolving market.