Bring the transformative power of the NVIDIA Blackwell architecture directly to your office environment. The Exxact Valence NVIDIA DGX Station™ redefines desktop computing by integrating the NVIDIA GB300 Grace Blackwell Ultra Superchip—a breakthrough design that couples a 72-core NVIDIA Grace™ CPU with two Blackwell Ultra GPUs via the ultra-fast NVLink-C2C interconnect. This unified superchip architecture eliminates traditional bottlenecks, providing a coherent memory pool of nearly 800GB to handle trillion-parameter models that previously required massive server clusters.

Unmatched Performance for Generative AI

Designed for the most demanding AI workflows, the Exxact Valence NVIDIA DGX Station™ delivers unprecedented speed and efficiency:

• Unified Memory Architecture: Access 496GB of LPDDR5X CPU memory and 288GB of HBM3e GPU memory as a single coherent pool, enabling you to fit massive LLMs and datasets entirely in memory.
• Extreme Compute Power: Achieve up to 20 petaFLOPS of AI performance with fifth-generation Tensor Cores and support for the new NVFP4 precision, accelerating training and inference by orders of magnitude over previous generations.
• Data Center Connectivity: Equipped with the NVIDIA ConnectX-8 SuperNIC, this station supports up to 400Gb/s networking, ensuring seamless integration with DGX SuperPODs and cloud clusters for scalable workflows.
  
  

Whisper-Quiet, Office-Ready Design

Despite its server-grade capabilities, the Exxact Valence is engineered for the human workspace. Its precision-tuned liquid cooling system ensures the GB300 Superchip operates at peak performance while remaining whisper-quiet (<35dB). No specialized power or cooling infrastructure is required—simply plug it into a standard wall outlet to turn your desk into a world-class AI research center.

Enterprise-Grade Software Stack

The system comes pre-installed with NVIDIA DGX OS and the comprehensive NVIDIA AI Enterprise software suite. From the moment you power on, you have instant access to optimized frameworks, pretrained models, and development tools like NVIDIA RAPIDS™ and TensorRT™, allowing you to go from prototype to production faster than ever before.

Optimized for Massive Scale and Future-Proof Workflows

The Exxact Valence is not just a standalone workstation; it is a gateway to massive-scale AI factories. The system is designed with architectural continuity, ensuring that models developed locally on the desktop can be seamlessly migrated to larger data center clusters like the NVIDIA GB300 NVL72 rack-scale system without code changes or optimization penalties. This capability allows researchers to prototype on the desktop and scale instantly to the cloud or data center.

Advanced Connectivity and Expandability

Beyond its core processing power, the workstation features robust I/O capabilities to support diverse professional needs:

• High-Speed Storage: Includes 2x PCIe 5.0 x2 M.2 NVMe slots (from CPU) and 2x PCIe 6.0 x4 M.2 slots (for training data), providing ultra-fast data access to feed the hungry GPUs1.
• Display Versatility: Supports a double-wide PCIe 5.0 x16 slot dedicated for display output, allowing you to drive high-resolution monitors for visualization and data analysis alongside heavy compute tasks.
• Networking Flexibility: Comes equipped with dual 10GBASE-T Ethernet ports and two QSFP112 ports supporting up to 400Gb/s via the NVIDIA ConnectX-8 SuperNIC, ensuring high-throughput connectivity for distributed training and large file transfers.
  
  

Precision-Engineered for Efficiency

Built to maximize performance per watt, the system leverages the NVIDIA Grace CPU's power efficiency combined with the Blackwell Ultra GPU's high compute density. The GB300 Superchip supports dynamic power management, and the system is housed in a sleek full-tower chassis (23.4" L x 9.6" W x 20.8" H) that fits comfortably in a standard office setting, powered by a single 1600W 80 PLUS Titanium power supply.