At AMD’s highly anticipated ‘Advancing AI’ event, CEO Lisa Su introduced a series of powerful new GPUs and processors, aimed at revolutionizing AI workloads and accelerating innovation in the sector. With the launch of its 5th generation EPYC processors and Instinct MI325X GPUs, AMD is positioning itself as a leader in AI infrastructure.
“Our latest Zen 5 Core is designed to be the best for server workloads, delivering 177% higher IPC than Zen 4,” said Su, underlining the processing power of the new EPYC chips. The new generation boasts up to 192 cores and 384 threads, promising to enhance both cloud and enterprise workloads. “We’ve built the broadest portfolio of CPUs, covering everything from new cloud workloads to important enterprise workloads,” Su explained.
AMD’s AI-focused hardware doesn’t stop at the CPU. Su announced the launch of the MI325X Instinct accelerator, which offers unparalleled AI performance.
“MI325X leads the industry with 256 GB of ultra-fast HBM3E memory and six terabytes per second of bandwidth,” Su stated, which underscores its potential to enhance generative AI performance. She also detailed its efficiency in running large models, promising between 20–40% better inference performance on key models like Llama.
Looking ahead, Su gave a preview of AMD’s upcoming MI350 series, set to launch in 2025.
“CDNA 4 delivers over seven times more AI compute and brings the biggest generational leap in AI performance in our history,” she remarked This shows confidence in AMD’s future growth in AI-driven sectors.
AMD’s latest innovations reflect its ambition to dominate the AI space by delivering cutting-edge solutions that cater to the increasingly demanding needs of enterprise and cloud environments.
