Traditional factories run on fragmented systems. As manufacturing enters the era of agentic automation, facilities need a new compute layer to handle real-time reasoning and massive sensor ingestion right on the floor.
To provide this physical foundation, ASRock Rack introduces the 4UXGM-GNR2 CX8 centralized server and the 2UXGI-THOR edge platform. Designed with advanced thermal engineering and high-throughput networking, these systems can support NVIDIA Factory Operations Blueprint (FOX) , a reference design for building an industrial “AI brain” that orchestrates fleets of industrial AI agents to resolve issues quickly.
Engineering the Infrastructure for Agentic Factories
Handling massive AI workloads requires specialized, purpose-built hardware. ASRock Rack provides solutions for both the data center and the factory edge.
Centralized AI Simulation: The 4UXGM-GNR2 CX8
Heavy data center orchestration happens centrally. To support this, the ASRock Rack 4UXGM-GNR2 CX8 acts as the primary compute hub. It is a 4U NVIDIA MGX™ server featuring a robust dual-socket architecture to support high-performance processors. The system includes a powerful 3200W CRPS power supply to handle intense processing demands. For rapid data transfer, it utilizes eight QSFP112 (400Gb/s) ports powered by NVIDIA ConnectX®-8 technology.
Distributed Edge Intelligence: The 2UXGI-THOR
On the factory floor, the environment becomes much more constrained. The ASRock Rack 2UXGI-THOR is a 2U platform built on the NVIDIA IGX Thor™ platform. It features IGX Thor module and RTX Blackwell GPU, delivering 5584 FP4 TFLOPS AI compute with rigorous vertical thermal engineering to maintain optimal operating temperatures, which prevents clock throttling during sustained generative AI tasks. To see exactly how this hardware manages these intensive localized workloads, you can read more about how the ASRock Rack 2UXGI-THOR enhances edge AI capabilities.
Supporting the Autonomous Factory
With the physical framework established, these systems can seamlessly run the most advanced industrial software available today, including NVIDIA FOX Blueprint that leverages open Nemotron models to create a real-time operational super agent.
The centralized 4U servers support NVIDIA RTX PRO™ 6000 Blackwell Server Edition GPUs. This setup delivers the raw compute performance necessary to process complex simulations. To connect the central servers with the edge devices seamlessly, the infrastructure relies on NVIDIA Spectrum-X™ Ethernet networking and NVIDIA BlueField® DPUs. This ensures secure, scalable communication across the entire facility.
The Paradigm Shift: Moving to Plant-Wide Intelligence
Implementing this hardware changes how a manufacturing facility operates fundamentally. In the past, rigid automation meant disconnected sensors and high latency. Maintenance was entirely reactive. Today, multi-agent orchestration becomes possible. Factories can process real-time sensor data, manage autonomous operations efficiently, and connect live machine signals into a unified decision layer.
Strategic Use Cases for Industrial AI Agents
These advanced capabilities translate into direct operational benefits across the manufacturing process.
- Automated Model Training: Using the ASRock Rack 4U server alongside the NVIDIA FOX Blueprint, developers can fine-tune AI models centrally with Nemotron or Cosmos libraries and skills and deploy them to the factory floor.
- Visual Operations and Compliance: The 2UXGI-THOR edge server ingests live, uncompressed video feeds. Utilizing NVIDIA Metropolis VSS Blueprint, developers can build real-time visual AI agents for applications like defect detection, process compliance checks, operational intelligence, and worker safety.
- Real-Time Operational Twins: Because of the high-speed networking, factories can visualize their live data using NVIDIA Omniverse™ libraries. This creates accurate, proactive digital representations of their physical operations.
Conclusion
Ultimately, achieving true factory intelligence requires the right physical equipment. ASRock Rack provides the thermal engineering, server design, and high-speed connections needed to deploy ruggedized platforms for industrial settings and running accelerated AI models such as NVIDIA FOX. By working with NVIDIA’s open models and accelerated platforms, we supply the essential infrastructure to build tomorrow's intelligent, automated factories.
