tldr: Dell Technologies and NVIDIA are showcasing joint innovations at SC25 to accelerate AI and HPC outcomes. The Dell PowerEdge XE8712 delivers industry-leading GPU density and the industry’s first intelligent, automated rack-scale platform for HPC, paired with PowerCool liquid cooling for thermal stability and efficiency. Dell Enterprise SONiC, our open-source network operating system, now supports NVIDIA Spectrum‑X, enabling a unified, lossless Ethernet fabric for AI and HPC environments.
The demand for AI and high-performance computing (HPC) is growing at an incredible pace. Industries from healthcare to finance are using HPC and AI models to solve complex problems and create new opportunities.
Dell Technologies, working with NVIDIA, is advancing high-performance computing (HPC) and AI at SC25 by announcing the availability of the Dell PowerEdge XE8712, the industry’s first intelligent and automated rack-scale system purpose-built for HPC, and Dell Enterprise SONiC support for NVIDIA Spectrum-X™ Ethernet for seamless networking.
Unmatched density and scalability
The Dell PowerEdge XE8712 is purpose-built for complex HPC simulations. The PowerEdge XE8712 is built for density, allowing maximum computational power in a minimal footprint.

-
- Rack-scale architecture with the NVIDIA GB200 NVL4 platform, featuring GB200 Grace Blackwell Superchip configurations designed for acceleratedcompute.
- Factory-tested and pre-configured Dell IR7000 Integrated Racks support up to 36 nodes with up to per rack—among the highest densities available—so you can run more training and simulation jobs in less space.
- Designed to reduce operational overhead through rack-level serviceability and quick disconnect manifolds, enabling efficient maintenance and faster scale-out.
Superior cooling technology
Intense workloads generate significant heat, which can limit performance and increase operational costs. The XE8712 features PowerCool liquid cooling innovation, including the new Dell PowerCool Rack-mount Coolant Distribution Unit (RCDU), which ensures stable thermals and consistent performance. Advanced PowerCool technology protects your infrastructure investment while maximizing efficiency.
Our cooling solutions include:

-
- Dell PowerCool RCDU: This unit delivers up to 160 kW of liquid cooling capacity, stabilizing thermals even under the most demanding AI and HPC workloads.
- Advanced leak detection: Intelligent management and leak detection through the Integrated Rack Controller (IRC) and OpenManage Enterprise provide rack‑scale visibility and risk mitigation across cooling components to reduce downtime and safeguard your hardware.
- Energy efficiency: By managing heat effectively, PowerCool liquid cooling technology helps lower energy costs by enabling higher facility water temperatures and reducing energy use compared to traditional air cooling.
Unified networking for HPC and AI
Powerful servers are only part of the equation. A high-speed, low-latency network is essential to ensure that GPUs are fed with data continuously. Any bottleneck in the network can lead to idle resources and slower outcomes.

- Enterprise SONiC Distribution by Dell Technologies, our open-source-based network operating system, now seamlessly supports NVIDIA Spectrum-X Ethernet platforms. As part of the PowerSwitch family of switches and providing a single point of support, this combination creates a powerful, unified networking fabric designed specifically for the demands of AI and HPC.
-
- Spectrum-X Ethernet combines Spectrum-4 switches with BlueField®-3 SuperNICs™ to deliver lossless Ethernet, dynamic load balancing, and congestion control—purpose-built for multi-tenant, GPU-intensive AI workloads.
- Dell SONiC capabilities for AI include RoCEv2, Priority Flow Control (PFC), and Enhanced Transmission Selection (ETS) to maintain deterministic performance and minimize tail latency in data-intensive training.
- Together, Dell SONiC + NVIDIA Spectrum-X Ethernet simplify fabric operations while maximizing throughput for distributed training—helping reduce idle GPU time and shorten training cycles.
This integrated solution offers low-latency, lossless networking that is critical for distributed training and other data-intensive tasks. Optimized for AI, it simplifies network management while delivering the performance needed to keep your GPUs operating at peak capacity.
Advancing progress together
The collaboration between Dell Technologies and NVIDIA is about more than just hardware. It’s about providing integrated, end-to-end solutions that empower organizations to innovate and succeed. From advanced servers like the PowerEdge XE8712 to a robust networking foundation, we are committed to helping you build the future. By bringing together best-in-class technologies, we help you accelerate research, develop smarter products, and solve some of the world’s most pressing challenges.
Ready to redefine HPC and AI? Learn more about Dell PowerEdge accelerated servers and Dell SONiC software to see how we can help accelerate your goals today.
Your guide to Dell AI Factory starts here.
Frequently Asked Questions (FAQ)
What makes the PowerEdge XE8712 ideal for AI and HPC?
The XE8712 offers industry-leading GPU density, advanced liquid cooling, and smart rack-level automation for unmatched performance and scalability. This combination allows for faster AI model training and more complex simulations while maintaining operational efficiency.
How does Dell SONiC support NVIDIA Spectrum-X Ethernet?
Dell SONiC integrates seamlessly with NVIDIA Spectrum-X Ethernet, offering a unified, low-latency, and lossless networking fabric designed for AI environments. This ensures data flows efficiently between GPUs, maximizing performance and reducing training times.
How can businesses benefit from this technology?
Businesses can achieve shorter AI training cycles, which accelerates time-to-insight and innovation. They can also realize operational savings from energy-efficient cooling and benefit from simplified management, empowering them to focus on strategic goals rather than infrastructure complexity.


