Computex 2024: Silicon Power Showcasing Industrial Storage Solutions for AI Applications

As the world transitions into the AI era, Silicon Power Computer & Communications Inc. exhibited a number of new industrial storage solutions featuring high performance, large capacities, and wide temperature ranges at Computex 2024 to empower the development of AI applications.

RAID solution for GPU direct storage architecture

The GDS3M0 is a RAID solution that eliminates the bandwidth constraints of traditional PC systems. Equipped with 8xPCIe Gen 4×4 SSDs in 1xRAID card, it eliminates transmission bottlenecks between GPUs, expanding the bandwidth by 4x, and improving latency by a 74% during sequential RW operations. It also delivers performance with sequential RW speeds up to 28GB/s and 26GB/s, ⁽¹⁾ respectively. The GDS3M0 ensures processing efficiency to smoothly handle the data-intensive workloads of AI applications.

Server storage solution: SSU3H0S U.2 SSD

Converting large amounts of data into commercially valuable information requires storage devices to facilitate data transmission and computing. The SSU3H0S provides RW performance reaching up to 7,000MB/s and 6,800MB/s, (1) respectively. With a maximum capacity of up to 30.7TB, it supports SATA, SAS, and PCIe interfaces, allowing enterprise customers to flexibly expand their storage infrastructure. Additionally, the SSU3H0S features hot-swappable capabilities, enhancing system stability.

Extended wide temperature: -40°C to 105°C

The company highlighted 3 types of DRAM (RDIMM, ECC SO-DIMM, ECC U-DIMM) capable of withstanding an industry emperature range of -40°C to 105°C. This extended temperature resistance makes these products for maintaining optimal performance and reliability inside AI-driven servers and systems, which can get very hot during the vast amounts of data processing and calculating. Additionally, the firm’s M.2 PCIe SSDs and industrial-grade portable SSD feature a standard wide temperature resistance range of -40°C to 85°C, ensuring they remain efficient in demanding environments.

Large capacities for high-volume AI-Driven workloads

The company offers high-capacity storage solutions throughout its complete product portfolio to meet the ever-increasing demands for more space. The firm’s M.2 PCIe SSDs, including the M.2-2280 with capacities up to 4TB and the M.2-2230 with capacities up to 1TB, are designed to meet the storage needs of AI applications. Large capacities are crucial for AI workloads, which require extensive data storage and rapid access to large datasets for training and inference processes. Additionally, the use of pSLC technology enhances endurance and reliability, ensuring that these SSDs can handle the rigorous demands of continuous RW cycles in AI environments.

With a focus on applications such as AI computing, smart healthcare, embedded systems, and industrial automation, the company will continue to expand its product portfolio and enhance its presence in the industrial storage sector, offering enterprise customers an even wider array of solutions.

^[1] Results are based on data from Silicon Power’s internal lab. Actual results may vary based on the environment, usage patterns, or other factors.