MSI has introduced the XpertStation WS300, a desktop workstation engineered specifically for artificial intelligence computing. The compact system combines a 72 core NVIDIA Grace processor with Blackwell Ultra graphics and an enormous 748GB memory configuration. However, official retail pricing and shipment schedules remain unannounced.
The internal hardware targets massive machine learning workloads directly at the desktop level. The platform links the central processor with graphics silicon to access 496GB of LPDDR5X system memory alongside 252GB of high speed HBM3e memory. According to official performance figures shared by NVIDIA, this setup delivers up to 20 PFLOPS of FP4 processing power. That provides enough local compute capacity to run complex models with up to 1 trillion parameters without relying on external cloud clusters. For local storage, the workstation includes 2 NVMe drives offering 2TB each, configured in a mirrored RAID 1 array to protect operating system files.
System expansion remains an option for enterprise users who need extra graphical horsepower. A dedicated PCIe slot allows builders to install an extra NVIDIA RTX Pro Blackwell graphics card. For larger network environments, the system supports an NVIDIA ConnectX 8 SuperNIC adapter capable of data transfer rates up to 800 Gbps. This allows multiple WS300 units to link directly into a shared processing cluster. A 1600W Platinum power supply handles the electrical draw, while a custom liquid cooling loop manages thermal output during sustained heavy computation.
