The world's first 2nm GPU is here! Su Zifeng launched the "most powerful AI rack", and the CPU performance surpassed that of NVIDIA

Zhidong Author | ZeR0 Editor | Mo Ying Zhidong reported in San Jose on July 23 that just now, at the annual AMD Advancing AI Conference, AMD Chairman and CEO Dr. Su Zifeng made a series of announcements, releasing AMD's strongest AI chip Instinct MI455X GPU, "the strongest intelligent CPU" Venice, "the strongest rack-level AI infrastructure" Heilos, and a Ryzen chip specially designed for robots. Overall, AMD announced 7 major releases today: Rack-level AI platform Heilos Server CPU "Venice" Instinct MI400 series GPU (including MI455X and MI430X) Air-cooled PCIe GPU MI350P AI software stack ROCm.ai Local AI developer platform "Gorgon Halo" Kria AI robot system module & developer platform At the same time, Su Zifeng announced AMD The latest roadmap for GPU, CPU, and rack-level AI infrastructure. Among GPU products, the MI500 series is expected to be released next year, using the next generation HBM, supporting larger vertical expansion networks, and introducing new copper interconnect and optical interconnect technologies; the MI600 series is under development and will be released in 2028. The inference throughput of the next generation MI500 will soar to more than 2000 times that of MI300X. Among server CPUs, the sixth-generation "Venice" released today is AMD's first 2nm server CPU, using 512 threads and PCIe 6.0 for the first time; the seventh-generation server CPU "Florence" will use a new generation of process nodes and memory and is expected to ship in 2028; the eighth-generation "Ravenna" is under development and is expected to ship in 2030. In the rack AI infrastructure, the roadmap for the past three years is as follows: Helios was released today, using "Venice" CPU, MI455X series GPU, Pensando "Vulcano" A