Dell announced a suite of new products and services—many intended to address roadblocks to companies adopting generative AI and LLMs—at the International Conference for High Performance Computing, Networking, Storage, and Analysis, held from Nov. 17 to 22 in Atlanta.
Varun Chhabra, senior vice president of Infrastructure and Telecom Product Marketing at Dell, said the most pressing issues preventing AI adoption are quality of data, cost, and energy, power demands, and sustainability concerns.
“Enterprises are pursuing AI to remain competitive in today’s digital landscape, but they need to harness their proprietary data to differentiate,” Dave Vellante, chief analyst at theCUBE Research, said in a press release. “We’re seeing a resurgence in interest for offerings like Dell’s AI infrastructure, solutions, and services that can help customers prepare their data to extract insights, automate tasks, and transform processes.”
Dell announced three new server products: the PowerEdge XE7740 and PowerEdge XE9685L, and updates to the Integrated Rack 5000 (IR5000) series.
The PowerEdge XE7740 with Xeon 6 processors is appropriate for inference and fine-tuning. Eight front PCIe slots allow for high-speed network connections for GPU-to-GPU performance.
“This is probably the best platform for enterprise adoption to start,” Chhabra said.
The Dell PowerEdge XE9685L holds up to 96 NVIDIA Tensor Core H200 or B200 GPUs per rack for high-density computing needs, such as AI training.
“Not only are we building the platform, we expect that we can deliver all of this in a rack-scale solution out-of-the-factory integrated,” Chhabra noted.
Both PowerEdge products, which will be released in the first quarter of 2025, are available through the Dell Integrated Rack Scalable Systems program.
Relatedly, Dell also debuted a turnkey solution in the Dell Integrated Rack 5000 series, an expansion of the rack 7000 series, which can be either liquid-cooled or air-cooled and can hold 96 GPUs per rack. It will be compatible with the upcoming NVIDIA GB200 Grace Blackwell NVL4 Superchip.
Also revealed at the conference, Dell Data Lakehouse will now support unified access control between the Trino query engine and the Apache Spark processing system.
“Dell does the homework and the legwork of testing and validating the most common ecosystem options customers have for their AI workloads,” Chhabra said.
“The continued reimagining and improvement of AI and HPC technologies is critical to accelerating discoveries across industries,” Dan Stanzione, executive director of the Texas Advanced Computing Center, said in a press release. “With Dell’s AI solutions, we are building complex HPC systems that allow innovation to thrive in research communities. Ultimately, these advances have the potential to reshape society and drive progress that benefits everyone.”
In the realm of AI infrastructure, Chhabra said Dell is dedicated to optimizing its AI Factory by leveraging NVIDIA hardware. At the Supercomputing event, Dell revealed they’ll offer support for NVIDIA Tensor Core GPUs by the end of the year. This includes the H200 SXM cloud GPUs.
“We spend a lot of time working with NVIDIA on integrating their software upgrades, testing and validating their software upgrades on top of our infrastructure,” Chhabra said.
Dell plans to support the NVIDIA H100 NVL GPU in upcoming PowerEdge servers and offers a new solution for general professional users and data scientists: Dell Agentic RAG with NVIDIA. Its RAG agents handle company-specific data access and regulatory challenges. These are available now, while the GPU update will be available before the end of the year.
Lastly, Dell announced a handful of services available as of Nov. 18:
“As we continue into this new era of technological innovation,” Vellante explained, “it will be critical for enterprises to continuously learn and invest in innovation to realize the benefits AI promises.”
