SK Hynix PS1010 E3.S Enterprise NVMe SSD

The SK Hynix PS1010 is a purpose-built PCIe Gen5 x4 NVMe SSD in the compact E3.S 1T form factor, designed to maximize storage density and throughput within modern AI server and hyperscale data-center architectures. Built on SK Hynix's advanced 238-layer TLC NAND and paired with an in-house enterprise-class NVMe controller, the PS1010 delivers sequential read speeds up to 14 GB/s and random read performance exceeding 2.5 million IOPS, enabling it to feed high-bandwidth AI inference pipelines and large-scale dataset retrieval operations without becoming an I/O bottleneck. The E3.S form factor is specifically optimized for sled-based and Open Compute Platform (OCP) server designs, offering superior thermal management through its flat, ruler-style profile that facilitates front-to-back airflow in high-density 1U and 2U chassis configurations.

The PS1010 incorporates end-to-end data protection mechanisms including T10 DIF/DIX (Protection Information), hardware-accelerated AES-256 encryption compliant with TCG Opal 2.0 and TCG Enterprise SSC, and power-loss protection (PLP) circuitry backed by onboard capacitors to ensure in-flight data is safely committed during unexpected power events. The drive supports the NVMe 2.0 specification with multiple namespaces, NVMe-MI (Management Interface) for out-of-band telemetry and health monitoring, and Zoned Namespace (ZNS) readiness, making it compatible with modern software-defined storage stacks and AI/ML data pipeline orchestration frameworks.

Launched in 2025-Q3, the PS1010 targets enterprises building out GPU server clusters for AI training and inference, cloud service providers scaling read-intensive object storage tiers, and HPC deployments requiring consistent ultra-low latency at queue depth 1. With capacities spanning up to 15.36 TB in a single E3.S module and a rated endurance profile suitable for mixed enterprise workloads, the PS1010 represents SK Hynix's commitment to delivering vertically integrated NAND-to-controller solutions that reduce dependency on third-party components while maintaining strict quality and reliability standards for mission-critical environments.

Ideal for

• AI inference acceleration: serving as the high-speed local storage tier for GPU server nodes hosting large language models and multimodal AI inference engines requiring rapid random read I/O
• Hyperscale read-intensive object storage: acting as the primary storage layer in software-defined object storage clusters where high random read IOPS and density per rack unit are critical SLAs
• High-performance computing scratch storage: providing ultra-low-latency temporary dataset storage for HPC workloads processing large scientific, genomic, or financial simulation datasets
• AI training data pipelines: staging preprocessed training datasets close to GPU compute nodes to minimize data starvation during iterative deep learning model training runs
• Virtualization and cloud-native workloads: underpinning NVMe-oF storage fabrics or local NVMe pools in hypervisor environments running thousands of concurrent virtual machines or containerized microservices
• Database acceleration: supporting OLTP and in-memory database logging tiers where sub-100-microsecond read latency and sustained mixed random I/O throughput directly impact transaction rates

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.