Samsung PM9E1: Inside the Leading AI-optimized PCIe Gen5 SSD

As AI processing capabilities continue to evolve toward on-device and desktop supercomputer applications, the need for ultra-fast yet space-efficient local storage becomes increasingly important. Samsung’s recent innovation in AI-optimized storage devices drives this point home. Our PM9E1 SSD ― the industry’s first PCIe Gen5 4TB SSD in M.2 22x42 form factor ― complements the NVIDIA GB10-powered NVIDIA DGX Spark™↗ desktop AI supercomputer.

While AI accelerators enable systems to harness the vast power of AI, only pairing them with performance-optimized, purpose-built storage technology like the Samsung PM9E1 can provide the massive capacity and blazing access speeds to unleash that power that may lead to the development of tomorrow’s killer AI apps.

Power and performance redefined for AI

The Samsung PM9E1 boasts twice the performance characteristics with up to 45% enhanced power efficiency over previous generation devices.¹

• Strategic hardware design with dual-sided PCB integration of 8th-generation 1Tb V-NAND and DRAM, enabling high capacity and consistent performance in a compact form factor
• Sequential Read and Write speeds of up to 14,500MB/s and 12,600MB/s
• Random Read and Write speeds of up to 2,000K IOPS and 2,640K IOPS

But what makes it AI-centric? AI processing requires a hybrid of both the immediacy of real-time processing and the ability to handle bulk data transfers rapidly. The storage system must not only have the ability to deal with both, but also excel at both.

PM9E1 uses an in-house ‘Presto’ controller design, built on 5nm Samsung Foundry technology with firmware optimized for DGX Spark OS software, NVIDIA CUDA↗, and the overall AI user experience.

Even in the compact M.2 22x42 form factor, PM9E1 integrates a full performance architecture with dedicated DRAM, enabling consistent high-speed data access and low-latency response. This allows the drive to efficiently handle both large sequential transfers and intensive random access patterns required by AI workloads, ensuring responsive performance for time-sensitive AI tasks.

Built-in security for the AI era

What good is high performance if the transactions are susceptible to attacks? Using significant AI processing power just to verify transaction security would be self-defeating.

Validated to all the latest security standards, the PM9E1 controller includes security protocol and data model (SPDM) v1.2 support for device authentication, firmware attestation, and secure channels. This version of SPDM provides for certificate-based security, bringing the same tight level of authentication and support for encryption used for external communications down to the component data interchange level. 

Qualified and in mass production for the recently introduced desktop AI Supercomputer: NVIDIA DGX Spark

The high-performance specifications of PM9E1, near bullet-proof operation and reliability, and transaction security support the requirements of NVIDIA DGX Spark desktop AI supercomputing.

SSD speed and capacity may be the stars of the show, but achieving these in a low-power and compact design is what enables an entire AI supercomputer to fit into such an appealing desktop package. SSD power efficiency reduces thermal load, creating greater flexibility in system layout. The M.2 22x42 form factor brings the storage medium physically closer to compute resources, making the most of limited space while also allowing the fastest possible signal paths. 

Conclusion

Considering its powerful combination of performance, capacity, and security characteristics, PM9E1 occupies a unique position among mass storage options ― allowing it to unleash innovation and creativity in local on-device AI workstation applications. Its advanced design means that these workstations, and the creators who use them, can be confident that they have a long runway before reaching the limits of the technology.

Samsung storage delivers future-proof system design through forward-looking features that prepare customers for the next era of computing. Its selection for use in the upcoming generation of desktop AI supercomputers and creative workstations testifies to the thoroughness of its design and qualification, and makes it the obvious choice to support the AI compute of today and tomorrow.