close
close

Dell refines AI factory and expands Nvidia partnership

LAS VEGAS — Dell is leaning on its concept of the AI ​​factory, a stack of technologies aimed at addressing AI training and workloads in the enterprise.

At Dell Technologies World 2024, the supplier unveiled an expanded partnership with Nvidia. The two companies first launched the Dell AI Factory with Nvidia at GTC in March. Here, Dell refines the Nvidia partnership with new integrations and offerings. It is also refining its contributions to the AI ​​Factory with new AI PCs and workstations, as well as networking and server products, while expanding its partner ecosystem with Hugging Face, Meta and Microsoft Azure to accelerate the training and deployment of AI workloads .

CEO Michael Dell kicked off the conference with a history lesson about factories – from the early factories that kept the wheels turning to the different ways to automate processes. Electricity allowed for faster automation and machines for even more. Now factories are going one step further, Dell told attendees.

“We are moving from computation to cognition to the age of AI,” he said.

We are moving from computation to cognition to the age of AI.

Michael DellCEO, Dell

Dell was joined on stage by Nvidia CEO Jensen Huang, ServiceNow CEO Bill McDermott and Samsung SDS CEO Sungwoo Hwang. They talked about how their core technologies are now colored by AI and automation.

“Every workflow and every enterprise in every industry and corner of the world will be reinvented with GenAI,” said McDermott.

A closer look

The Dell AI Factory is a set of curated and purpose-built technologies for AI. It includes a new portfolio of Copilot+ AI PCs with on-device access to Copilot and powered by Qualcomm Snapdragon Series X; new storage for structured and unstructured data and an increase in capabilities for the PowerEdge XE9680 server, which Dell called the heart of the AI ​​factory.

On the storage front, Dell has added a new PowerScale scalable device, the PowerScale F910 all-flash array, which the company says offers a 127% performance increase over the F900 and better density. For mission-critical data such as databases, Dell also unveiled PowerStore Prime, a unified storage offering, during the conference.

“Data was and is at the center of everything. It’s the rocket fuel. So let’s go where the data is. And no company in the world has delivered more storage capacity than Dell,” Dell said.

Dell has expanded its networking portfolio to support AI workloads with a new Dell PowerSwitch Z9864F-ON, powered by Broadcom’s Tomahawk 5 chipset for AI workloads with high throughput and low latency requirements.

The PowerEdge XE9680 server now supports Broadcom’s 400G PCIe 5.0 Ethernet adapters for improved performance. Dell is also overlaying its SmartFabric Manager for SONiC software on top of its hardware to simplify deployment and management for customers.

While most data center workloads are still considered outdated today, the advent of AI could change the way a data center operates, according to Dave Vellante, an analyst at theCUBE Research. He noted that Huang suggested a year ago that virtually all workloads would be accelerated by AI, eventually transforming data centers into AI factories.

“The AI ​​factory is a three-layer vision: infrastructure that is AI-enabled; automation, so you can take labor out of the picture as much as possible; and tooling, which includes things like large language models and LLM options,” says Vellante. said.

AI applications and services would be layered across the infrastructure of an AI factory, Vellante said. Dell does not have AI applications to offer, but will work with partners and provide services and infrastructure to help customers build, deploy and run AI apps. One of those partners is Hugging Face, to create Dell Enterprise Hub on Hugging Face. The hub allows customers to train LLMs on Dell infrastructure, within their data centers.

Parallel File System: A New Frontier for Dell

The vendor also highlighted Project Lightning, its proprietary parallel file software, at Dell Technologies World. This is designed for PowerScale and aimed at supporting AI workloads.

According to Henry Baltazar, an analyst at 451 Research, part of S&P Global Market Intelligence, AI workloads have become so large, mainly due to unstructured data, that users are starting to see similar problems as with supercomputing. Struggling with large data sets and the significant computational requirements for data processing are some of the drivers behind the use of parallel file systems, which can be difficult to use and are now moving to the cloud for AI workloads.

“Dell needs to respond,” he said. “That’s why we’re seeing this kind of technology being developed.”

While Dell is working to provide a dedicated AI factory offering with Nvidia, the company says Dell AI Factory technologies could work with other AI accelerators, including those from AMD and Intel.

AI Factory with Nvidia is growing

The Dell AI Factory with Nvidia takes the concepts of the AI ​​Factory, but only builds them with Nvidia hardware and software.

Dell is expanding its server options with the XE9680L which comes with direct liquid cooling, which helps increase GPU density per server. The cooling allows for up to eight Nvidia Blackwell GPUs in a 4U form factor or 72 in a rack, which Dell says is the highest in the industry.

The Dell NativeEdge, an edge orchestration platform first introduced last summer, now automates Nvidia AI business software. Dell is also adding new services with Nvidia, including Dell Generative AI for Digital Assistants, to accelerate the deployment of these bots.

Dell AI Factory with Nvidia provides full stack automation for rapid deployment of AI environments. And Dell Accelerator Services for RAG on Precision AI Workstations enables customers to use fetch-enhanced generation on a Dell Precision workstation with Nvidia AI Workbench, a development platform.

Nvidia’s current dominance in AI is comparable to the duopoly of Microsoft and Intel in the 1990s and early 2000s, Vellante said. Dell then went big on partnerships and is now doing so again with Nvidia.

“That’s important because if you want to be an AI vendor, you’re going to have to have GPUs in your stack,” he said.

Adam Armstrong is a TechTarget editorial writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.com.