AMD boosts AI performance in new line of enterprise PC chips

The neural processing unit performance of newer chips opens a new front in the battle with Intel and Qualcomm for AI desktop supremacy.

AMD’s big Advancing AI event in San Francisco on Thursday underlined how quickly the microprocessor industry has pivoted to artificial intelligence (AI) as its main sales pitch.

The company offered three hardware announcements across its processor line-up, each appealing to different parts of the AI market.

The first was the new Instinct MI325X AI accelerator chip, a datacenter-oriented GPU which ups performance on every metric compared to last year’s MI300. The company also showed off its fifth-generation EPYC processors for the enterprise cloud and datacenter sector. And it unveiled the new Ryzen AI PRO 300 series, a family of processors for mobile PCs aimed at enterprise buyers.

Mobile chips have traditionally been low power (and lower performance versions) of their desktop equivalents, but with the focus on AI, that distinction is fast disappearing.

AI requires more raw power, which is now showing up in the specifications for new chips. For example, the Ryzen AI PRO 300 series offers three processors, starting with the Ryzen AI 9 HX PRO 375; it features 12 Zen 5 cores/24 threads, a clock speed that can be boosted to 5.1GHz, and integrated Radeon 890M graphics.

It also features a neural processing unit (NPU) that delivers up to 55 tera operations per second (TOPS), making it the most powerful mobile NPU of its kind on the market.

Just below that in the line-up is the Ryzen AI 9 HX PRO 370, an identical chip with slightly less NPU performance — up to 50 TOPS. And the entry-level chip is the Ryzen AI 7 PRO 360 with 8 cores/16 threads, a 5GHz clock speed, Radeon 880M graphics, and the same 50 TOPS NPU performance.

NPUs are on the new frontline of desktop competition because they make possible features such as accelerating Microsoft’s Copilot + and AI-intensive tasks such as real-time language translation.

They also pit AMD against traditional rival Intel, which has had NPUs of its own in its Core Ultra CPUs since 2023. Today, NPU AI acceleration is premium priced, but there are signs the technology is likely to jump quickly to more mainstream chips.

“With Ryzen AI, we’ve actually enabled hundreds of different AI functions,” said AMD CEO Lisa Su near the end of a two-hour Advancing AI presentation. “Our latest software stack makes it really easy for developers to optimize thousands of pre-trained models for Ryzen.

“Our Ryzen AI Pro 300 series resets the bar for what a business PC can do,” she said.

AI-enabled chips could quickly become standard issue across all sectors. Earlier this year, IDC predicted that by 2027, PCs with AI acceleration built in would constitute 60% of all PC shipments.

On the same day as AMD’s event, Intel announced its new Core Ultra 200S processors, which the company claims deliver NPU performance of 36 TOPs between its CPU, GPU and NPU. A third player in the market offering NPU technology is Qualcomm, which launched its Snapdragon X PC chip in September.

News Sources: Computerworld

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top