Intel’s new chip Gaudi3 competes with Nvidia and AMD

On Thursday, Intel launched a new range of computer chips, including Gaudi3 – an AI chip specifically designed for generative AI software applications. Set to launch next year, Gaudi3 will compete with leading chips from Nvidia and AMD that currently power large, energy-intensive AI models.

Big players in AI models, such as OpenAI’s ChatGPT, utilize Nvidia GPUs hosted in the cloud, which is a contributing factor to Nvidia’s nearly 230% rise in stock year to date, compared to Intel’s 68% rise. This is the reason why companies like AMD, and now Intel, are introducing their chips in hopes of luring AI companies from Nvidia’s commanding market position.

Even though specific details are scarce, Gaudi3 is set as a rival to Nvidia’s H100 and AMD’s upcoming MI300X, which are the go-to for firms constructing vast chip farms for AI applications. It is slated to start shipping to customers in 2024.

Intel has been developing Gaudi chips since acquiring chip developer, Habana Labs, in 2019.

Intel CEO Pat Gelsinger said at a launch event in New York where he announced Gaudi3 along other chips focused on AI applications; “We’ve been seeing the excitement with generative AI, the star of the show for 2023.”

“We think the AI PC will be the star of the show for the upcoming year,” Gelsinger added. And that’s where Intel’s new Core Ultra processors, also announced on Thursday, will come into play.

Intel Core Ultra and new Xeon chips

In addition to Gaudi3, Intel also introduced new Core Ultra chips, which are intended for Windows laptops and PCs, as well as the fifth-generation Xeon server chips. Both kinds of chips incorporate a specialized AI component referred to as an NPU, designed to expedite the execution of AI programs.

This perhaps signals an industry-wide pattern where traditional chip manufacturers, including Intel’s competitors like AMD and Qualcomm, are reconfiguring their product lines and alerting shareholders to the potential increase in demand for their chips, driven by AI models.

Intel’s newly introduced Core Ultra may not enable the powering of an AI chatbot like ChatGPT offline, but it’s capable of managing smaller tasks. As an example, Intel mentioned that Zoom utilises their chips to operate its background-blurring feature. These chips are constructed using Intel’s 7-nanometer process, offering greater power efficiency than previous iterations.

Significantly, the introduction of these 7-nanometer chips implies that CEO Gelsinger’s goal to match Taiwan Semiconductor Manufacturing Co.’s chip manufacturing expertise by 2026 remains on track.

The Core Ultra chips also come with enhanced gaming features and added graphical power. This can help software, such as Adobe Premier, to achieve over 40% boosted performance. These chips were launched in laptops that became available in stores on Thursday.

Also, Intel’s fifth-generation Xeon processors are typically used to power servers run by larger entities such as cloud companies. Intel did not announce the pricing for these, but its predecessor was priced in the thousands. These Xeon processors are often paired with Nvidia GPUs in systems used for training and implementing generative AI, some systems even combining eight GPUs with one or two Xeon CPUs.

Intel indicated that the latest Xeon processor will be particularly effective for inferencing, or deploying an AI model, a process which requires less power compared to the training process.