Intel Unveils New AI-capable CPU

A new Intel microprocessor due later this year will have the ability to run a generative AI chatbot on a laptop without relying on the cloud, says the chip giant at its annual Innovation conference this week.

Intel CEO Pat Gelsinger is plotting a comeback not only by leveraging new chipmaking techniques, but also by incorporating AI capabilities into its latest chips.

AI on your laptop

In his vision, AI use won’t be confined to data centers or the public clouds run by the cloud giants, where they rely heavily on chips made by Nvidia. Instead, AI use will spread to new areas, including the PC market.

This means users will potentially be able to run large language models and other machine learning workloads such as Llama 2 and Stable Diffusion without the need for a dedicated GPU from Nvidia or AMD. Running everything locally offers a level of privacy and control that cloud-based systems cannot deliver.

As reported by The Register, most of the demos at the conference highlighted Intel’s upcoming 7nm Core Ultra processors, code-named Meteor Lake. The chips are among Intel's first client processors to feature a new multi-die architecture pairing an Intel CPU die with a GPU made by TSMC.

Crucially, the microprocessors feature an integrated neural processing engine (NPU), which is designed specifically to speed up the performance of common machine learning tasks.

With the NPU, developers working with a new version of Intel’s OpenVINO software will be able to run a version of large language models. Interestingly, Intel is also extending support for OpenVINO to ARM processors, which is used predominantly by smartphones.

Intel says this is to support AI inference on low-power edge systems, including IoT sensors that are powered by ARM-compatible controllers and chipsets.

“AI is fundamentally restructuring science and so many domains, unleashing new applications and new experiences in productivity and creativity. But we also believe that ushers in the next era of the PC: the AI PC generation,” Galsinger said.

Paul Mah is the editor of DSAITrends. A former system administrator, programmer, and IT lecturer, he enjoys writing both code and prose. You can reach him at [email protected].​

Image credit: iStockphoto/hapabapa