Supermicro Introduces Super HPC Station Powered by the NVIDIA GB200 Developer Kit: Data Center Performance in a Deskside Form Factor

Exclusive collaboration delivers deskside platform for HPC application development

Supermicro announced today the Super HPC Station powered by the NVIDIA GB200 Developer Kit, a deskside computing platform that delivers data center-grade performance for HPC workloads directly to developers' workspaces. Based on the NVIDIA GB200 Grace Blackwell Desktop Platform, this developer-focused system enables engineers to develop, run, and test high-performance computing (HPC) applications locally with the same architectural capabilities found in production data centers.

Data Center-Grade Performance in a Developer-Accessible Form Factor

The Super HPC Station addresses a fundamental challenge in HPC development: providing developers local access to the same architecture and performance characteristics they'll encounter in production deployment such as NVIDIA GB200 NVL72. The system provides up to 682 GB of coherent memory, FP64 compute capabilities, and the performance required for developing and testing large-scale HPC workloads in a deskside environment.

Built for HPC Application Development

The system integrates the NVIDIA GB200 Grace Blackwell Desktop Platform, combining a powerful NVIDIA Blackwell generation GPU featuring NVIDIA CUDA cores and NVIDIA Tensor Cores, with the NVIDIA Grace CPU. The NVIDIA NVLink-C2C interconnect between these components enables high-bandwidth, low-latency communication essential for memory-intensive HPC applications.

Key technical specifications include*:

*Preliminary specifications, subject to change.

Flexible Deployment Options

The Super HPC Station offers flexible form factor options to match different workspace requirements. Available in both deskside and racked configurations, the system can be deployed wherever developers need immediate access to compute resources. The plug-and-play design enables teams to get started quickly without extensive infrastructure setup.

Closed-loop direct-to-chip liquid cooling for the CPU, GPU, NVIDIA ConnectX-8, and memory ensures quiet operation, making the system well-suited for deskside deployment in research and development environments. This thermal solution maintains optimal performance while minimizing acoustic impact.

The NVIDIA ConnectX-8 SuperNIC provides up to 800 Gb/s of networking bandwidth, enabling multi-system configurations and seamless integration into existing development environments.

Large Coherent Memory for AI Workloads

Modern AI models demand increasing memory capacity as they grow in scale and complexity. The GB200 desktop platform powering the Super HPC Station addresses this requirement with a coherent memory architecture that allows large-scale models to operate within a single memory space. By eliminating traditional memory hierarchy constraints between CPU and GPU, the C2C interconnect provides high-bandwidth coherent access across processors and accelerators, enabling efficient training and inference for memory-intensive workloads.

Exclusive Availability

The Super HPC Station powered by the NVIDIA GB200 Developer Kit is exclusively available through Supermicro, with targeted availability at the end of Q1 2026. This collaboration combines Supermicro's expertise in high-performance system design with NVIDIA's leadership in accelerated computing.

For more information about the NVIDIA GB200 Developer Kit, visit https://www.nvidia.com/en-us/products/workstations/gb200-developer-kit/

The NVIDIA GB200 Developer Kit accelerates HPC and AI workload development by providing data center-level capabilities in a flexible, deskside form factor. Targeted availability: End of Q1 2026.