Dell adds to its AI Factory, extending its services and hardware to cover new AMD-based servers in its push to provide ways to run AI workloads on premises.
Dell is adding five AMD-based servers to its AI Factory, a portfolio of infrastructure and services for AI workloads.
The new PowerEdge servers include four fifth-generation AMD EPYC Turin processor models: the R6715 and R7715 single CPU as well as the R6725 and R7725 dual CPU models. The fifth server, which also utilizes the new AMD EPYC Turin CPUs, is the XE7745 that can house up to eight double-width GPUs or 16 single-width GPUs and is designed for AI inferencing.
Along with new hardware, Dell also released version 10 of its integrated Dell remote access controller server management software, bringing more security through its 256-bit encryption. Also part of the expanding AI Factory is the new Dell Generative AI Solutions with AMD, which combines infrastructure, software and professional services. Dell also expanded its partnership with Hugging Face, an open source platform for model training, to include AMD-based hardware so that customers can choose what infrastructure to build their models on.
Dell's investment in AMD makes sense as it offers higher performance than Intel CPUs at a lower cost, according to according to Anurag Agrawal, CEO of Techaisle, a global SMB and IT channel market research and analyst firm. It also offers better performance per watt over Intel CPUs.
"Whether you're a mid-size or enterprise customer, you are looking to optimize costs and looking at ROI on AI," Agrawal said.
While the new AMD servers expand the options for hardware in Dell's AI Factory offering, the vendor still needs to add clarity around how customers can best use it, Agrawal said. The Dell AI Factory provides building blocks and steps for AI but not steps with a clear path.
"Factory concept is, 'I go with the raw material, and I come out with an end product,'" he said.
AMD adds more building block options, but the lack of clarity on how to go from raw material to product remains, he said.
Servers and inferencing
The four new AMD-based R-series servers are designed with data center modular hardware systems that allow for better air cooling for the use of dual 500W AMD CPUs.
Whether you're a mid-size or enterprise customer, you are looking to optimize costs and looking at ROI on AI.
Anurag AgrawalCEO, Techaisle
Camberley Bates, an analyst at the Futurum Group, said this new architecture is a significant step forward for both vendors in terms of energy efficiency and density. The new servers don't just lower CPU costs but also lower the energy costs per performance, she said.
"We know that data centers are struggling with the power and cooling requirements for AI deployments," she said. "Thus, this has become a competitive position for the server vendors."
Dell stated that the new servers offer record breaking performance in VMmark 4, a measure of virtualization performance, and TPCx-AI, a measure of machine learning benchmarks, making them ideal for AI and analytics. But they can also be used for databases.
The PowerEdge XE7745 is designed for AI inferencing and model tuning. The XE, or extreme acceleration, series is aimed at data intense workloads including AI and high-performance computing. The XE9680 was released in the second half of 2023. While the 9680 relies on Intel CPUs, the 7745 uses AMD CPUs and accelerators, and it includes GPU slots that are paired with PCIe 5.0 slots for network connectivity.
The XE7745 is aimed at on-premises AI workloads. Regulated customers, such as those in financial services or healthcare, can't go to the cloud to do inferencing as use cases come up, according to Agrawal. Dell wants to be ready as the use cases arrive on premises, he added.
"The use case story is still being told," Agrawal said.
Professional services expand to AMD
Beyond hardware and software, the Dell AI Factory offers professional services such as Kubernetes configurations, AI deployment frameworks and best practices for ROI. Dell has now expanded this to include AMD environments as well.
Dell's overall go-to-market strategy is customer choice, Agrawal said.
"Their professional services organization does this Accelerator Workshop and then they give them the give them a choice, and the choice is based on a cost and deployment time frame," he said.
While the PowerEdge XE9680 supported AMD accelerators since its release, the hardware now includes other aspects of the AI Factory, such as Dell Generative AI Solutions with AMD and the Dell Enterprise Hub on Hugging Face, a portal that enables customer to use Hugging Face on Dell equipment. The Hugging Face-Dell portal will now support the server with AMD Instinct MI300X accelerators for model deployment on premises.
"Overall, this is part of Dell's strategy to streamline and make it easy to deploy the systems on premises," Bates said.
Adam Armstrong is a TechTarget Editorial news writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.
