Intel launches AI chips for servers and clients

Intel launched a series of chips for AI processing at its AI Everywhere event in New York, reinforcing its plans to power AI workloads that span the data center, cloud, network and PCs.

The 5th Gen Intel Xeon Scalable processor, codenamed Emerald Rapids, comes just one year after Intel launched its fourth generation of Xeon Scalable, known as Sapphire Rapids. Intel says that compared with the previous generation of Xeon, these processors deliver an average 21% performance gain for general compute performance and 36% higher performance per watt, on average, across a range of workloads.

For now, Xeon is the only mainstream data center processor with built-in AI acceleration, which Intel claims delivers up to 42% higher inference and fine-tuning on models as large as 20 billion parameters over the prior generation of Xeon. They come with up to 64 cores and fit in the same thermal envelope as the 4th Gen Xeon.

Intel said customers following a typical five-year refresh cycle, and upgrading from even older generations of Xeon chips can reduce their TCO by up to 77%.

Outside of AI, the new Xeon architecture will benefit other use cases as well, Intel says. At the event, IBM announced that 5th Gen Intel Xeon processors achieved up to 2.7 times better query throughput on its watsonx.data platform compared to previous-generation Xeon processors.

And Google, which intends to deploy 5th Gen Xeon next year in its cloud service, said that Palo Alto Networks experienced a two-fold performance boost in its threat-detection deep learning models by using built-in acceleration over the previous generation of Xeon processors.

Also at the event, Intel showed off its Gaudi3 dedicated AI processor, saying it is on track to ship sometime in 2024.

On the client side, Intel showed off the Core Ultra mobile processor family, which is the first built on the Intel 4 process technology and the first to utilize AI processing on the client.

AI on the client is important because it will complement and support server-side AI processing, most notably in the area of inference. Some data cannot be processed in the cloud, such as regulated data, and must remain within the confines of a company’s firewall. So to protect the integrity of data and take the load off the servers, inference processing is moving to the client.

Intel CEO Pat Gelsinger said in a statement: “Intel is developing the technologies and solutions that empower customers to seamlessly integrate and effectively run AI in all their applications — in the cloud and, increasingly, locally at the PC and edge, where data is generated and used.”

Francis Sideco, principal analyst with TIRIAS Research, attended a briefing with Gelsinger, and the issue of the new chips coming so soon after the fourth generation was brought up. Gelsinger responded, “what do you want me to do, slow down?”

Sideco couldn’t disagree with Gelsinger on that one. “They want to get these out and get caught back up in leadership. The only way he felt they could get back to the leadership that they were, that they’re shooting for, is to go fast,” he said. Sideco noted that AMD is also on a fast release cadence with its Epyc server processors.

Servers with 5th Gen Intel Xeon Scalable processor will be available beginning in the first quarter of next year.