Micron HBM3E integrated into AMD instinct MI350 series for AI and HPC workloads

Micron Technology, Inc. announced that its HBM3E 36GB 12-high memory is integrated into the AMD Instinct MI350 Series GPUs and platforms. This memory solution supports the training of large AI models, high-throughput inference, and complex high-performance computing (HPC) tasks, including data processing and computational modeling.

The Micron HBM3E 36GB 12-high memory provides high bandwidth and reduced power consumption for the AMD Instinct MI350 Series GPU platforms, which are based on AMD’s CDNA 4 architecture. These platforms feature 288GB of HBM3E memory capacity, delivering up to 8 TB/s bandwidth. The Instinct MI350 Series GPUs can support AI models with up to 520 billion parameters on a single GPU. In a full platform configuration, they offer up to 2.3TB of HBM3E memory and achieve peak theoretical performance of 161 PFLOPS at FP4 precision. The integration of Micron’s HBM3E with AMD’s architecture supports large language model training, inference, and scientific simulations, enabling data centers to scale compute performance.
The collaboration between Micron and AMD has focused on optimizing compatibility between the HBM3E 36GB 12-high product and the Instinct MI350 Series, reducing time to market for AI s...