Server consolidation gets new life as agent AI puts a strain on CPUs

Server consolidation gets a second look as agent AI triggers a surge in demand for general-purpose computing. Graphics processing units dominate the AI ​​infrastructure conversation, but the agents running on these systems rely heavily on central processing units, and data centers don’t have enough power or space to add more servers.

Dell Technologies Inc. sees this pressure as an opportunity for its PowerEdge server line, which expanded with 11 new servers in May. According to Pradyumna Narendra (pictured), product planner and senior consultant at Dell, replacing aging fleets with denser systems creates capacity for new AI work.

“As AI becomes faster, reasoning and the agent AI surrounding it will also accelerate and grow faster,” Narendra said. “As these use cases scale, the computational needs grow exponentially.”

Narendra spoke with John Furrier of theCUBE Research in the AI ​​Data Pipeline Forum during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed how server consolidation, rack-level efficiency, and support from multiple chip suppliers help customers make room for agentic AI. (*Disclosure below.)

Server consolidation creates space for agent computing

Dell’s upcoming 18th generation PowerEdge servers are based on this update math. Systems powered by Advanced Micro Devices Inc. processors will deliver 70% performance gains from generation to generation and can combine 13 Dell 14th-generation servers into one, Narendra said. Profits are amplifying as companies retire older hardware.

“For every 10 to 13 servers you replace, you get a lot of that productivity improvement back because you can now do that with one server,” Narendra said. “If you now extend this to all data centers, it is like an acceleration effect that you can benefit from.”

Pure processor speed is only part of the argument, as many server manufacturers use the same AMD silicon. Dell’s three-rack, two-socket design, for example, uses less power for cooling than a two-unit equivalent, allowing customers to pack more computing power into each rack, he explained.

“So by using servers that are slightly larger in package, you get higher density at the rack level, which sometimes seems counterintuitive,” Narendra said. “But we are actually seeing results and our customers are benefiting from them.”

Management tools like the Integrated Dell Remote Access Controller (iDRAC) and OpenManage Enterprise tie these fleets together, while support from multiple chip suppliers prevents server consolidation from becoming a forced decision. Narendra says customers get systems that look and function the same regardless of what processors are inside.

“We believe in democratizing technology and enabling silicon diversity so that our customers are not locked into a single operating model or choice,” Narendra said. “That is the promise and the premise with which we have built our portfolio.”

Here is the full video interview, part of SiliconANGLE and theCUBE’s coverage of the AI ​​Data Pipeline Forum:

(*Disclosure: Dell Technologies Inc. sponsored this segment of theCUBE. Neither Dell Technologies Inc. nor other sponsors have editorial control over the content of theCUBE or SiliconANGLE.)

Photo: SiliconANGLE