Micron’s 9550 PCIe Gen5 E1.S data center SSDs are making waves in the AI world by earning a coveted spot on NVIDIA’s Recommended Vendor List (RVL) for their NVIDIA GB200 NVL72 AI system. This powerful combination delivers a compelling solution for AI workloads, enabling efficient performance and energy savings. The 9550 SSD’s inclusion in the GB200 NVL72, powered by the GB200 Grace Blackwell Superchip, represents a significant step toward optimizing AI infrastructure at a rack scale. This integration fosters energy-efficient AI deployments, making it an ideal solution for tasks like large-scale AI model training, real-time trillion-parameter language model inference, and demanding high-performance computing (HPC) operations.
Micron’s 9550 SSD stands out with its remarkable performance advantages. When paired with NVIDIA Magnum IO GPUDirect (GDS), the 9550 SSD delivers up to 34% higher throughput compared to other industry offerings. This translates to a significant boost in efficiency, particularly when handling demanding workloads like graph neural network (GNN) training with Big Accelerator Memory (BaM), where completion times are up to 33% faster.
Beyond its performance, the 9550 SSD prioritizes sustainability by setting new benchmarks for energy consumption. Micron’s commitment to environmental responsibility is evident in the 9550 SSD’s ability to consume 81% less SSD energy per 1TB transferred than other SSD offerings with NVIDIA Magnum IO GDS. This energy efficiency extends to GNN training with BaM, where the SSD boasts up to 43% lower SSD power consumption.
Jeremy Werner, Corporate Vice President and General Manager of Micron’s Storage Business Unit, highlights the crucial role of Micron’s memory and storage solutions in meeting the evolving demands of AI workloads. “Micron’s memory and storage products play a critical role in meeting the growing requirements of demanding AI workloads from the data center to the edge. By integrating the Micron 9550 SSD on the GB200 NVL72, server companies can integrate a high-performance, energy-efficient Gen5 data center storage solution into their AI server systems,” Werner said.
Keith Morris, Vice President of Product Management at NVIDIA, emphasizes the importance of NVMe storage for the GB200 NVL72 rack-scale design. “Ultra-fast and energy-efficient NVMe storage is crucial to the NVIDIA GB200 NVL72 rack-scale design. The Micron 9550 SSD can be integrated by our solution partners into their systems based on the GB200 NVL72 reference architecture to enable higher performance and efficiency,” Morris said.
The Micron 9550 SSD, in collaboration with NVIDIA’s GB200 NVL72, is paving the way for a more sustainable and efficient future of AI computing, driving innovation in areas like large-scale AI model training, real-time language model inference, and high-performance computing. This partnership promises to unlock new possibilities for businesses and researchers seeking to harness the power of AI while minimizing their environmental footprint.