On April 17, the Solidigm team visited LUISUAN Technology for in-depth technical exchanges. This exchange focused on storage solutions in the AI ​​era. The two parties had heated discussions around the technological development of hard drives, G3.5 storage application scenarios, and high-performance computing storage platforms, laying a solid foundation for potential future cooperation.

On Tao Storage: A New Solution to QLC in the AI ​​Era

At the exchange meeting, the Solidigm team first shared its latest technical parameters and development routes for QLC and TLC hard drives, and elaborated on the balance strategy between capacity, performance, durability and cost. LUISUAN Technology also provided professional insights on the application prospects of QLC disks in G3.5 storage scenarios. The two parties exchanged views on how to optimize the storage architecture in high-density and high-efficiency AI workloads.

Subsequently, Kong Weihai, general manager of LUISUAN Technology Products and R&D Center, introduced the company's development history and product system to the guests, focusing on the GP Spark series of high-performance storage platforms. This series includes the GP Spark 2000 storage companion for desktop-level CMX applications, and the GP Spark 3000 platform with integrated switching and storage capabilities, which has aroused strong interest from the Solidigm team.

GP Spark: a storage engine for AI

LUISUAN Technology’s GP Spark series is specially designed for AI computing and edge reasoning scenarios and has the following highlights:

GP Spark 3000：It is not only a desktop CMX-level switching storage platform, but also integrates 4×100Gbe or 8×50Gbe switching capabilities, supports 8 E1.S PCIe 5.0 SSDs, and can achieve ultra-low latency data transmission through RDMA technology. Its "cable direct connection expansion" design supports direct connection of multiple DGX Spark devices, greatly simplifying the deployment and expansion of small and medium-sized AI clusters.

GP Spark 2000：As a "storage companion", it achieves enterprise-class storage performance in a compact size. A single machine supports four NVMe SSDs, with a maximum capacity of 248TB per disk. It has the characteristics of plug-and-play and seamless integration with the Spark kernel level. It is suitable for a variety of flexible scenarios such as edge AI inference nodes and desktop-level AI development environments.

Technological resonance: jointly exploring a new blue ocean of AI storage

The Solidigm team has shown high recognition for the GP Spark series, especially in terms of its integrated design integrating storage and network switching, low-latency access capabilities, and modular expansion solutions. Both parties agreed that with the trend of AI computing power penetrating to the edge and desktop, storage systems not only need higher bandwidth and IOPS, but also need to be simple, efficient and easy to expand in architecture.

This exchange was not only a collision of technical viewpoints, but also an important dialogue on the evolution of AI infrastructure between two companies with deep accumulation in the fields of storage and computing. In the future, the two parties are expected to carry out more in-depth cooperation in storage media optimization, system architecture collaboration, scenario-based solutions, etc., and jointly promote the innovation and implementation of the AI ​​storage ecosystem.