AMD Debuts New DPUs to Boost AI Workload Efficiency

At its Advancing AI event, Advanced Micro Devices (AMD) showcased its next generation of artificial intelligence (AI) chips.

During a press briefing, AMD’s SVP and general manager of the network technology solutions business, Soni Jiandani, stated that the new DPUs will aid in addressing network issues brought on by AI workloads.

The statements made at an AMD presentation in San Francisco come as demand for AI processors from major technology companies like Microsoft and Meta Platforms vastly outstrips supply.

AMD revealed its fifth-generation EPYC server CPU, which CEO Lisa Su claims delivers significantly more power than its predecessor, as well as new Ryzen AI PRO processors for enterprise AI.

The business also unveiled its next Instinct MI300X family GPU, the MI325X, with production slated to begin in the fourth quarter and availability for AMD partners in the first quarter of 2025.

It claimed that its market share for EPYC server CPUs reached a new high of 34% at the end of the second quarter, and it expects to grow further with its next generation of products.

“Explosive new AI growth is putting new challenges on the network, requiring innovation at a much faster pace and infrastructure that has to be adaptable to meet with the growing needs of the traffic patterns of AI,” Jiandani said.

“It is absolutely critical to have a high-performing network along with a high-performing CPU, DPU, within an AI system. Our focus at AMD is to innovate in every aspect of the AI system and we believe that networking is not only critical, but it’s foundational to drive optimal performance.”

Analysts predicted that the event will serve as a “catch-up” catalyst for AMD to extend its share of the market, following a solid start to the year, with data centre sales reaching a record high in Q2.

We earlier reported that Samsung Electro-Mechanics (SEMCO) is going to supply AMD, NVIDIA’s primary rival in the artificial intelligence (AI) chip market, with high-performance substrates for hyperscale data centres.