ORICO Launches FlashMem FM9800 SSD with Dedicated pSLC AI Cache for Local Inference

ORICO launched the FlashMem FM9800 SSD, an M.2 2280 drive with PCIe Gen5 support and a dedicated pSLC AI Cache for storing local AI model weights.

ORICO Launches FlashMem FM9800 SSD with Dedicated pSLC AI Cache for Local Inference

ORICO introduced the FlashMem FM9800 solid-state drive on November 13 to address the growing storage demands of local artificial intelligence workloads. This drive targets creators and developers who run large language models on their desktops, as these applications require fast access to model weights and intermediate data. The design shifts focus from raw sequential speed to the latency needed for AI inference tasks.

ORICO targets local AI workloads with a new M.2 SSD featuring a dedicated pSLC cache area for model weights and inference data.

The FlashMem FM9800 operates as an M.2 2280 NVMe drive that supports the PCIe Gen5 interface. It utilizes 3D TLC NAND flash memory to store data, which provides a balance of capacity and endurance for modern systems. The drive integrates a graphene thermal pad to manage heat dissipation during sustained high-speed transfers.

Specifications

  • Form Factor: M.2 2280
  • Interface: PCIe Gen5
  • NAND Type: 3D TLC NAND
  • Cache Technology: pSLC AI Cache
  • Capacity Options: 768GB + 85GB and 1638GB + 128GB

A key technical feature of the FM9800 is its pSLC AI Cache configuration. ORICO allocates a portion of the NAND capacity to operate in pSLC mode, creating a dedicated high-speed buffer. This cache is specifically designed to store AI model weights, context , and local inference intermediate data, which reduces latency for AI applications.

The product launches in two specific capacity configurations to accommodate different storage needs. The first option provides 768GB of main storage paired with an 85GB cache area. The second option offers 1638GB of main storage with a larger 128GB cache area for more extensive model libraries.

Comments (0)

Add a comment

Keep it civil, relevant, and useful to other readers.