Dell PowerEdge XE8712 Advanced research data center server launched at GTC 2025 for High-performance AI and HPC workloads, maximum density and efficiency.
At NVIDIA GTC 2025, Dell PowerEdge made interesting announcements about NVIDIA Blackwell Ultra, NVIDIA RTX 6000, NVIDIA networking, and other products.
Advancing AI and Computing Power for Research
Dell is pleased to present the Dell PowerEdge XE8712, one of its most cutting-edge research data Centre server advancements to date, at GTC 2025. This server, which is intended to push the limits of massive AI and HPC workloads, continues Dell’s tradition of providing cutting-edge technology to businesses that want unmatched accelerated computing density, efficiency, and performance.
The PowerEdge XE8712 provides a precisely tailored platform to satisfy your demands, whether you’re training large AI models, simulating molecules, or making future predictions.
The PowerEdge XE8712’s integration with the NVIDIA GB200 NVL4 architecture, which was initially revealed at SC24, is its fundamental component and sets new standards for AI-powered rack solutions. The XE8712 has four NVIDIA NVLink-connected B200 GPUs, two NVIDIA Grace CPU Superchips, and a GB200 Grace Blackwell Superchip.
The PowerEdge XE8712’s Unique Features
AI and High Performance Computing (HPC) infrastructure are advanced by the PowerEdge XE8712 in the following ways:
- GPU Density Unmatched: The PowerEdge XE8712 has one of the largest GPU densities in the industry, supporting up to 144 NVIDIA Blackwell GPUs per Dell IR7000 series factory-integrated rack. This lowers operating expenses without sacrificing processing capacity by allowing more AI and HPC workloads to be completed in less physical space.
- Intelligent Cooling and Power: Utilizing cutting-edge Direct Liquid Cooling (DLC) technology, the XE8712 provides outstanding performance and can handle up to 264kW of power per rack. A 480kW shared power bus bar and disaggregated power shelves are elements of the XE8712’s flexible connectivity with the Dell IR7000. By effectively capturing heat output, its hybrid DLC system dramatically lowers energy usage and encourages environmentally friendly, sustainable operations.
- Adaptable Rack-Scale Architecture: With an IR7000 rack integration inspired by the Open Compute Project (OCP) Open Rack v3 (ORv3), the XE8712 offers remarkable scalability, enabling up to 36 nodes per rack. With Dell’s cutting-edge quick-disconnect DLC manifolds and a shared power bus bar, its front I/O serviceable architecture makes maintenance easier and provides seamless flexibility for changing infrastructure requirements.
Constructed for adaptability, this modular rack easily fits into dynamic AI environments and is prepared to handle next-generation Dell PowerEdge servers, meeting the increasing need for sophisticated cooling and power solutions.
Ideal Use Cases
The XE8712 is designed to tackle any task, including genetic sequencing, financial market modelling, and the application of AI in scientific research. The system’s capacity to provide dense compute solutions with enormous scalability, handling the most taxing workloads with remarkable efficiency, has already won over institutions engaged in spearheading large-scale research programs.
More Innovations from Dell at NVIDIA GTC 2025
Exciting NVIDIA GPU announcements for the factory clients are part of PowerEdge server portfolio expansion, in addition to the PowerEdge XE8712 joining the infrastructure options for the Dell AI Factory with NVIDIA.
Dell PowerEdge XE7740 & XE7745
Up to eight of the forthcoming NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs can be supported by the PowerEdge XE7745 and XE7740 servers, which offer outstanding performance and versatility. These servers are designed specifically for demanding AI, graphics, and compute-intensive workflows, and they are excellent at activities like model fine-tuning, AI inferencing, and new enterprise AI applications. They provide efficiency and adaptability for even the most complicated workloads to their sophisticated design.
Sneak Peek at NVIDIA Blackwell Ultra Platform
With the NVIDIA Blackwell Ultra platform, which has up to 288GB of HBM3e and is tailored for workloads like Large Language Model (LLM) training, Dell is preparing for the future. For a first look at the future NVIDIA HGX B300 NVL16 and NVIDIA GB300 NVL72 systems, which promise ground-breaking performance at almost infinite scale, be sure to stop by the Dell booth at NVIDIA GTC 2025.
Networking Breakthroughs with NVIDIA ConnectX-8
NVIDIA ConnectX-8 networking at 800 Gb/E is now supported by Dell’s PowerEdge servers. ConnectX-8, which promises peak performance even under high demand, serves as the backbone for data centers handling AI-driven workloads due to its ultra-low latency and expandable capacity.
