Dell unveils PowerEdge XE7740 server with Gaudi 3 AI accelerators

Dell Technologies has launched the PowerEdge XE7740, the first server to incorporate Intel's new Gaudi 3 PCIe AI accelerators for enterprise use.

The Dell PowerEdge XE7740 is positioned to provide scalable and cost-efficient artificial intelligence infrastructure to sectors including healthcare, finance, retail, and manufacturing. This server platform is structured to enable enterprise customers to integrate AI capabilities within existing data centre environments, focusing on flexibility, performance and ease of deployment.

Integrated AI accelerators

The PowerEdge XE7740 is the first server available to market to feature the Intel Gaudi 3 PCIe AI accelerator in an integrated configuration. The Gaudi 3 components are designed to deliver high performance for AI workloads, supporting up to eight accelerators in a single server. This scalability is expected to assist enterprises in managing large-scale AI models such as Llama 4, DeepSeek, Qwen and Falcon.

The 4U chassis PowerEdge XE7740 combines integration capabilities found in the PowerEdge R-Series with the scalability of the XE-Series. According to Dell, this model supports advanced networking configurations and powerful accelerator combinations to address the varying needs of contemporary AI development.

A Dell Technologies spokesperson stated, "The PowerEdge XE7740 delivers the ease of integration of the PowerEdge R-Series with the performance and scalability of PowerEdge XE-Series, advanced networking and powerful accelerator configurations to meet the growing needs of AI development."

Networking capacity and deployment

The XE7740 accommodates up to eight double-wide or PCIe accelerators, with configurations that allow two groups of four-way bridged accelerators. Bridging four Gaudi 3 PCIe accelerators achieves a combined throughput of 1,200 GBps for accelerator-to-accelerator communications over 18 x 200GbE interfaces using Remote Direct Memory Access (RDMA) over Converged Ethernet (RoCE v2). This level of accelerator-to-accelerator communication is critical for supporting larger models and data sets, facilitating workflow scalability and greater memory space availability.

It also features up to a one-to-one accelerator to network interface card (NIC) ratio through eight full-height PCIe slots and an OCP networking module. The XE7740 is designed to fit air-cooled server racks operating around 10kW, matching the power and cooling capacities common to many existing data centres.

AI model compatibility and workflow integration

The server is intended for seamless integration into enterprise environments. The XE7740 incorporates Dell Smart Cooling technology to optimise operation within air-cooled racks. It supports a range of popular AI models including Llama4, Llama3, Deepseek, Phi4, Qwen3, and Falcon3, ensuring broad compatibility for diverse AI workloads.

The Gaudi 3 accelerators are optimised to operate with mainstream AI development frameworks such as PyTorch, and connect readily with ecosystems including Hugging Face. This allows organisations to perform tasks such as large language model inferencing, fine-tuning, image and speech recognition, healthcare analytics, and real-time fraud detection within financial services.

The Intel Gaudi 3 PCIe accelerators integrate smoothly with PyTorch and platforms like Hugging Face, enabling organisations to fine-tune models or infer complex AI models efficiently.

Accessibility and efficiency

The XE7740's design focuses on removing common barriers to AI adoption in enterprise settings. Core benefits highlighted include cost efficiency and scalability, as well as compatibility with existing data centre infrastructure without requiring major power or cooling upgrades.

The server also aims to reduce operating costs associated with cloud-based AI processing and data egress. As described by Dell, "Cost efficiency: The XE7740 offers a compelling price-to-performance ratio, ensuring you get more AI capability for less. Scalability: With modular configurations and optional accelerator bridging, enterprises have greater flexibility to right-size and grow their AI capabilities incrementally. Compatibility: Designed for enterprise data centres with stringent power and cooling limitations, the XE7740 ensures seamless integration, eliminating the need for retrofitting infrastructure."

By providing a solution for on-premises AI infrastructure, enterprises can execute AI workloads while managing data security and compliance within internal networks.

Use cases and future applications

The Dell PowerEdge XE7740 is intended for a wide range of enterprise AI workloads. Examples include LLM inferencing, multimodal AI applications such as image and speech recognition, healthcare data analytics, fraud detection in financial services, and real-time personalisation for retail and eCommerce operations.

Looking towards the future, Dell positions the XE7740 as an adaptable component for evolving AI technologies. The company commented, "Dell PowerEdge XE7740 with Gaudi 3 is enabling more than today's AI workloads - it is about preparing enterprises for the future of AI while keeping a competitive edge. This scalable and adaptable infrastructure offers durable value, allowing businesses to pivot effortlessly as AI technologies evolve. With seamless integration into existing ecosystems, the XE7740 accelerates AI adoption and ensures businesses can meet their unique workload requirements with precision, staying ahead in an increasingly competitive business landscape."

The PowerEdge XE7740 server featuring Gaudi 3 accelerators is now available.