Specifications

Capacity

4 TB

Form Factor

M.2 22x42 (industry-first Gen5 4TB in this size)

Interface

PCIe 5.0 x4 NVMe

Sequential Read

14,500 MB/s

Sequential Write

12,600 MB/s

Random Read

2,000K IOPS

Random Write

2,640K IOPS

NAND

Samsung 8th-gen 1Tb V-NAND (dual-sided PCB with DRAM)

Controller

In-house "Presto" 5nm Samsung Foundry

Power Efficiency

+45% vs previous generation

Security

SPDM v1.2 · Device Authentication · Firmware Attestation

Qualification

NVIDIA DGX Spark (GB10) — qualified & mass production

Award

CES 2026 Innovation Awards Honoree

Overview

The Samsung PM9E1 is the industry's first PCIe Gen5 4TB SSD in the ultra-compact M.2 22x42 form factor, purpose-built for AI workloads. Combining Samsung's 8th-generation 1Tb V-NAND with the in-house "Presto" controller fabricated on a 5nm Samsung Foundry process, the PM9E1 delivers sequential reads up to 14,500 MB/s and writes up to 12,600 MB/s — with up to 45% higher power efficiency than its predecessor.

The PM9E1 has been specifically qualified and is in mass production for the NVIDIA DGX Spark desktop AI supercomputer powered by the GB10 processor. Its compact dual-sided PCB packs high capacity and full performance architecture into a minimal footprint, enabling AI supercomputers to fit into desktop-friendly packages while minimizing thermal load and signal path lengths.

Hardware-level security validated to the latest standards includes SPDM v1.2 support for device authentication, firmware attestation, and certificate-based secure channels. Firmware is optimized for DGX Spark OS, NVIDIA CUDA, and the overall AI user experience, making the PM9E1 the ideal storage foundation for AI workstations, creative systems, and next-generation on-device AI computing.

AI Workstation Integration

The PM9E1's dedicated DRAM cache ensures consistent low-latency data access under mixed AI workloads — handling both real-time streaming inference and bulk model training data transfers. Its compact 22x42 mm dimensions allow storage placement physically closer to compute resources, reducing latency in intensive I/O operations common in LLM inference, video generation, and scientific computing.

Request a Quote — Samsung PM9E1 4TB

QS Compute — NVIDIA Goldendisk Partner. AI-optimized storage for DGX Spark and workstation deployments.

Get Your Quote →