Solidigm D7-PS1010 7.68TB E3.S PCIe Gen5 SSD

The Solidigm D7-PS1010 7.68TB is a purpose-built enterprise NVMe SSD housed in the EDSFF E3.S 2T form factor, leveraging PCIe Gen5 x4 host connectivity to deliver sequential read throughput exceeding 14 GB/s and sequential write throughput approaching 7 GB/s. Designed from the ground up for next-generation AI server platforms, it combines Solidigm's 192-layer QLC NAND with an advanced NVMe 2.0-compliant controller that supports Namespace Management, ZNS optional features, and end-to-end data path protection via T10 DIF/DIX. The E3.S 2T mechanical footprint enables hot-swap serviceability, improved airflow management, and higher rack-level storage density compared to traditional 2.5-inch U.2 drives, making it an ideal building block for GPU server platforms and disaggregated storage nodes.

At 7.68TB raw capacity with an over-provisioned endurance profile rated at 1 DWPD, the D7-PS1010 is optimized for read-intensive and mixed-workload AI training dataset staging, model checkpoint storage, and high-throughput object storage back-ends. The drive implements full power-loss protection (PLP) via onboard capacitors, ensuring data integrity during unexpected power events—a critical requirement in enterprise and hyperscale deployments. Latency is tuned for queue-depth-heavy NVMe workloads, achieving sub-100 µs read latency at low queue depths and sustaining millions of IOPS at higher parallelism levels typical of AI accelerator-attached storage fabrics.

The D7-PS1010 supports NVMe over Fabrics (NVMe-oF) readiness, Firmware Update (FWUG), and Sanitize operations compliant with NIST 800-88 Rev1, addressing data security requirements for regulated enterprise environments. Its thermal design accommodates both active-cooled chassis with side-to-back airflow and conduction-cooled configurations increasingly common in liquid-assisted AI rack deployments. Solidigm's enterprise firmware stack provides robust telemetry via NVMe SMART and vendor-specific log pages, enabling integration with modern AIOps and predictive maintenance platforms.

Ideal for

• AI inference server local storage: staging large language model weights and KV-cache datasets directly attached to GPU nodes for low-latency, high-bandwidth model serving
• AI/ML training dataset repositories: hosting multi-terabyte training corpora in dense E3.S shelves connected to GPU clusters via NVMe-oF, eliminating network storage bottlenecks
• Hyperscale object storage nodes: deploying high-capacity QLC drives in software-defined storage clusters (Ceph, MinIO) to maximize usable petabyte density per rack unit
• High-performance database acceleration: supporting in-server NVMe tiering for OLAP and real-time analytics databases requiring sustained sequential scan throughput above 10 GB/s
• Model checkpoint and ephemeral scratch storage: providing fast intermediate storage for distributed training frameworks (PyTorch DDP, Megatron-LM) requiring frequent large sequential writes
• Next-generation disaggregated storage fabric: populating CXL- and NVMe-oF-attached JBOF enclosures in composable infrastructure architectures where E3.S hot-swap serviceability reduces operational overhead

Technical specifications

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