Seagate begins development on NVMe hard drives for cost-effective AI workloads

Seagate develops NVMe HDDs for AI tasks, improving cost-efficiency and performance.

: Seagate is revolutionizing HDDs by combining NVMe technology, aiming at cost-effective AI workloads in data centers. This development replaces proprietary interfaces with a simple PCIe protocol, enhancing compatibility and throughput. While NVMe HDDs won't match SSD speeds, they offer reduced carbon impact, lower power use, and reduced AI workflow latency. Seagate continues advancing its HAMR-based platform while collaborating on NVMe solutions to boost high-capacity drive capabilities.

Seagate is propelling hard disk drive (HDD) technology into the future by integrating it with NVMe (Non-Volatile Memory Express) technology, specifically targeting the needs of artificial intelligence (AI) applications. Despite HDDs traditionally falling short of the performance offered by SSDs (Solid State Drives), Seagate's approach aims to provide a more cost-effective and environmentally efficient solution for large-scale AI workloads. The move to a PCIe-based protocol allows HDDs to discard traditional SAS (Serial Attached SCSI) and SATA (Serial ATA) interfaces, leading to higher throughput and lower latency, which are crucial in data-intensive AI model training and inference operations.

By introducing NVMe compatibility, these hard drives can now be directly connected to GPUs (Graphics Processing Units), effectively bypassing the CPU-centric data routing that usually results in bottlenecks. This architectural change is aligned with the demands of AI models, which require seamless data access and processing capabilities to handle vast datasets efficiently. The adoption of NVMe over Fabrics (NVMe-oF) further enables HDDs to integrate into distributed, scalable storage systems, broadening the possibilities for data management in cutting-edge data centers.

Seagate has demonstrated a proof-of-concept integration showcasing the potential of NVMe HDDs that involved collaboration with Nvidia BlueField DPUs (Data Processing Units) and AIStore software. While the company has not disclosed specific performance data, it reports significant advantages for AI workflows. Key benefits highlighted include reduced latency in data access for AI tasks and minimized overhead associated with older SAS/SATA systems. These improvements are expected to amplify the efficiency and scalability of AI applications, particularly in expansive multi-rack storage clusters.

In addition to technical improvements, Seagate emphasizes the ecological and economic benefits of its NVMe HDDs. They are designed to be ten times more efficient regarding embodied carbon per terabyte of data and consume four times less power than their SSD counterparts. Moreover, the lower cost per terabyte provides a financially viable alternative to SSDs, especially for enterprises managing extensive AI data operations. These attributes underscore Seagate's commitment to creating sustainable technology solutions in line with global environmental goals.

Seagate is also advancing its HAMR (Heat-Assisted Magnetic Recording) technology, embodied in the Mozaic 3+ platform, to enhance the capacities and efficiency of HDDs further. The company intends to continue fostering partnerships that explore new storage solutions utilizing NVMe technology, driving the evolution of high-capacity hard drives. This initiative places Seagate at the forefront of meeting the storage demands of next-gen AI data centers, which require robust, versatile, and sustainable data handling solutions.

Sources: TechSpot, Seagate, Nvidia

data center