AMD introduced the Instinct MI430X accelerator featuring 432 gigabytes of HBM4 memory and 19.6 terabytes per second bandwidth as part of its expanding artificial intelligence hardware portfolio designed for high-performance computing environments. The processor incorporates next-generation CDNA architecture optimized for hardware-based double-precision floating point calculations, positioning the chip as a successor to the MI300A model deployed in the El Capitan supercomputer system.
The company plans integration across major research installations, including Discovery at Oak Ridge National Laboratory, which will pair the accelerators with Venice EPYC processors on HPE Cray platforms for large-scale model training and scientific computation. European exascale system Alice Recoque will similarly combine MI430X units with Venice chips using Eviden's BullSequana infrastructure.
AMD continues development on the MI455X accelerator intended to compete against offerings from graphics processor rival NVIDIA in the artificial intelligence computing sector.
The company plans integration across major research installations, including Discovery at Oak Ridge National Laboratory, which will pair the accelerators with Venice EPYC processors on HPE Cray platforms for large-scale model training and scientific computation. European exascale system Alice Recoque will similarly combine MI430X units with Venice chips using Eviden's BullSequana infrastructure.
AMD continues development on the MI455X accelerator intended to compete against offerings from graphics processor rival NVIDIA in the artificial intelligence computing sector.
