Dell Expands High-Performance Computing Portfolio for AI Workloads
Dell is enhancing its high-performance computing (HPC) portfolio for artificial intelligence (AI) workloads by introducing an AMD-powered server designed specifically to support large language models (LLMs). This addition to Dell's lineup complements its existing options powered by Nvidia, giving customers more choices for their AI infrastructure.
Dell PowerEdge XE9680: The Latest Server
The latest server from Dell, called the PowerEdge XE9680, comes equipped with eight AMD Instinct MI300X accelerators. These accelerators stand out for their impressive performance capacity, exceeding 21 petaFLOPS, and their 1.5GB of high-bandwidth memory (HBM3). These features make them particularly suitable for businesses that want to train and operate their own in-house LLMs.
Scalability and Connectivity
One of the standout features of Dell's new server offering is its scalability. Customers have the ability to scale the systems they deploy using the global memory interconnect (xGMI) standard. Additionally, AMD's GPUs can be connected over an Ethernet-based AI fabric using the Dell PowerSwitch Z9664F-ON. This release follows Dell's previous launch of a unit featuring Nvidia H100 GPUs.
Dell Validated Design for Generative AI with AMD
Dell is also introducing a new standard called Dell Validated Design for Generative AI with AMD. This standard provides a framework for organizations to run their own hardware and networking architecture for LLMs. It emphasizes the use of AMD ROCm powered AI frameworks, an open-source package that includes drivers, development toolkits, and APIs compatible with AMD Instinct accelerators. These frameworks support popular AI software like PyTorch, TensorFlow, and OpenAI Triton, all of which have native support on the PowerEdge XE9680 with AMD accelerators.
Dell's Commitment to Standards-Based Networking
Dell's strategy includes a focus on standards-based networking, as demonstrated by its membership in the Ultra Ethernet Consortium (UEC). Unlike Nvidia, Dell and AMD advocate for an open Ethernet for AI, enabling interoperability of switches from different vendors within the same system. Dell encourages businesses to adopt an open strategy that encompasses computing, fabric, and storage components essential for powering in-house generative AI models.
Availability
The new hardware and services that make up Dell's latest push in AI are expected to be available in the first half of next year. This development is part of Dell's broader effort to provide versatile and powerful solutions for businesses engaged in generative AI and other advanced computing tasks.
