Dell Technologies is expanding its open ecosystem for generative AI with the PowerEdge Server XE9680, featuring eight AMD Instinct MI300X accelerators.

Dell announced on the 7th that it is strengthening its 'Dell Generative AI Solutions' portfolio in collaboration with AMD.

Through this, Dell is building an open ecosystem and expanding the diversity of its processors, enabling each company to implement a simpler and more reliable AI strategy with customized solutions when applying AI to its unique data.

The Dell PowerEdge XE9680 server, Dell's fastest-growing solution in history, will support the soon-to-be-released AMD Instinct™ MI300X accelerator, providing a wide range of choices for AI performance.

The PowerEdge XE9680 server, equipped with the MI300X accelerator, supports high-performance capabilities to unlock the value of data and differentiate your business with custom large language models (LLMs).

Equipped with eight MI300X GPUs and 192GB of 5.3TB/s high-bandwidth memory (HBM3) per GPU, it delivers 1.5TB of sustained HBM3 capacity per server and 21 petaflops of FP16 performance, enabling enterprises to train larger models and occupy less data center space, lowering TCO and gaining a competitive edge.

In addition to supporting the AMD MI300X Instinct accelerator, Dell is introducing the Dell Validated Design for Generative AI with AMD as the next step in its Generative AI solution. This will enable businesses to more easily build trustworthy generative AI.

Dell and AMD's Validation Design for Generative AI will expand Dell's generative AI solution ecosystem by supporting the AMD ROCm™-based AI framework, which is scheduled to be released next year, and will also support several open source LLMs.

AMD ROCm™ is an open-source stack of drivers, development toolkits, and APIs for AMD Instinct accelerators.

Dell Validated Designs for Generative AI help customers easily build reliable generative AI platforms that meet their requirements without relying on guesswork when it comes to integration, performance, and sizing.

LLM can be easily utilized through open source AI/ML frameworks such as PyTorch, TensorFlow, ONNX-RT, JAX, and Triton, which are optimized for ROCm.

Dell Open Manage Enterprise software, included with the PowerEdge XE9680, simplifies and accelerates the deployment of generative AI infrastructure, while CloudIQ AIOps software provides intelligent automation and ensures security through integrated cyber resilience and comprehensive zero-trust implementation.

The Dell PowerEdge XE9680, equipped with eight AMD Instinct MI300X accelerators, supports a MI300X GPU cluster connected to an Ethernet-based AI fabric via a Dell validated GPU fabric equipped with the Dell PowerSwitch Z9664F-ON. xGMI (Global Memory Interconnect) technology spanning this cluster linearly scales training and inference for distributed generative AI, ensuring low latency.

As a member of the Ultra Ethernet Consortium, Dell is committed to advancing standards-based networking technologies and is uniquely positioned to provide open access to a full range of generative AI solutions, including servers, storage, and network fabric.

The Dell PowerEdge XE9680 with AMD Instinct MI300X accelerator and Dell and AMD validation designs for generative AI are expected to be available in the first half of 2024.

Kim Kyung-jin, President and CEO of Dell Technologies Korea, said, “Generative AI will act as the most effective accelerator of innovation in recent decades.” He added, “Who can better streamline processes and reduce costs in utilizing this technology will determine their competitive advantage in the new digital age, so Dell plans to take the lead in providing customers with a variety of choices through an open ecosystem.”