Dell Technologies and NVIDIA have joined forces to address the complexity of deploying infrastructure for generative artificial intelligence (AI). Building an IT infrastructure for deep learning and AI can be challenging, as the technologies and practices differ from traditional enterprise IT applications. Generative AI, in particular, places new demands on IT teams across various industries.
Generative AI requires expensive and power-hungry GPUs, often in large quantities. Moreover, aligning storage to cater to the data requirements of these GPUs involves adopting new technologies like NVIDIA’s GPUDirect. Additionally, the software stack for generative AI is distinct from mainstream enterprise IT.
To simplify the process, Dell and NVIDIA introduced Project Helix earlier this year, aiming to deliver full-stack solutions by leveraging Dell and NVIDIA infrastructure and software. Now, the companies have unveiled the first concrete outcomes from the project.
One significant development is the introduction of validated designs for inference systems based on NVIDIA accelerators and software. These validated designs provide IT organizations with a recipe-driven approach to building their infrastructure. Dell has a proven track record in enabling rapid technology adoption through validated designs for various applications, including analytics and high-performance computing.
For generative AI, Dell and NVIDIA have co-engineered the Dell Validated Design, which serves as a blueprint for infrastructure construction. It consists of pre-tested configurations combining Dell PowerEdge servers with the appropriate NVIDIA accelerators. The validated design supports several server options, such as the Dell PowerEdge XE8640, PowerEdge XE9680, and PowerEdge R760xa—all equipped with the latest Intel Xeon processors. Notably, the absence of an AMD option is surprising, given AMD’s success in the space. The servers support between four and eight NVIDIA Hopper H100 GPUs connected via NVLink technology.
In terms of storage, Dell offers support for its PowerScale filter and ECS object storage. Both systems integrate with NVIDIA’s GPUDirect, enhancing performance and reducing latency during data transfer to GPUs within the cluster.
While hardware plays a crucial role in generative AI, the software stack is equally important. Each validated design from Dell heavily relies on NVIDIA’s enterprise AI stack. NVIDIA AI Enterprise, a cloud-native suite of AI and data analytics software, is a key element of these designs.
In addition to infrastructure solutions, Dell also announced new Dell Precision AI Workstations. Popular among data scientists and AI researchers, these workstations come in various models featuring AMD Threadripper and Intel Xeon processors, coupled with up to four NVIDIA RTX 6000 GPUs. Interested customers can access the new workstation models through Dell’s regular sales channels starting in August.
Highlighting the significance of professional services, Dell expanded its offerings to include generative AI support. The company’s professional services team engages directly with customers throughout the AI implementation lifecycle. They help devise a generative AI strategy, identify high-value use cases, and provide full-stack implementation services. Even after deployment, Dell’s professional services offer operational efficiency, managed services, and staff training, further reducing the risk and complexity associated with generative AI projects.
Generative AI is poised to transform enterprises and revolutionize digital transformation. However, the perceived complexity and cost often deter IT organizations from pursuing such projects. Dell’s validated designs and professional services, coupled with NVIDIA’s hardware and software offerings, aim to address these challenges and simplify infrastructure deployment for generative AI. By leveraging these solutions, IT practitioners can mitigate risks and achieve successful outcomes.