HPC AI infrastructure convergence drives HPE strategy

HPC AI infrastructure is not any longer a parallel track to traditional computing. They’re merging right into a unified computing foundation that the supercomputing world has been constructing toward for years.

The identical infrastructure once reserved for national laboratory modeling and simulation has grow to be the inspiration for enterprise AI factories, and agents are already running scientific simulations autonomously, said Trish Damkroger (pictured), senior vice chairman and general manager for HPC and AI infrastructure solutions at Hewlett Packard Enterprise Co.

“Those things that we call supercomputers at the moment are the very same infrastructure that you just need for modeling and simulation that you just need for AI,” Damkroger said. “I’m finally seeing a number of this convergence, where you may have AI steering your HPC workflows, or agents running your simulations.”

Damkroger spoke with theCUBE’s Dave Vellante on the AMD Advancing AI event, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the deepening AMD and HPE partnership, liquid cooling as an enterprise imperative and why sovereign AI is expanding beyond national laboratories. (* Disclosure below.)

HPC AI infrastructure converges around liquid cooling, sovereignty and AMD partnership

HPE’s relationship with AMD encompasses essentially the most powerful HPC AI infrastructure deployments on the earth. The Frontier system at Oak Ridge National Laboratory, the primary to succeed in hyperscale, is an AMD and HPE platform, and HPE is under contract to deliver Oak Ridge’s next Discovery system, also AMD-based.  Lisa Su, president and CEO of AMD, announced each the EPYC 6 processor and AMD Instinct MI430X accelerator on the event, reinforcing an AMD and HPE partnership that Damkroger said extends from national labs to regulated industrial enterprises.

“We’ve got an incredible relationship with AMD,” Damkroger said. “The number two system on the TOP500 can also be an AMD system, El Capitan at Lawrence Livermore, where I got here from.”

As silicon power requirements proceed to soar, liquid cooling is becoming essential. HPE’s Cray heritage gives it a deep foundation in that space, and its GX5000 platform, a part of the Discovery system, is designed to handle high thermal design power parts while meeting European requirements for warm-water cooling at as much as 45 degrees Celsius. Damkroger said that the identical learning curve that shaped HPC AI infrastructure is now a complication for mainstream enterprise data centers and organizations that drag their feet will face mounting cost and density disadvantages as AI silicon continues to scale.

“You’d have one server in a rack should you’re going to air,” Damkroger said. “It’s going to get to the purpose where it’s going to be ridiculous.”

Here’s the entire video interview, a part of SiliconANGLE’s and theCUBE’s coverage of the AMD Advancing AI event:

(* Disclosure: TheCUBE is a paid media partner for the AMD Advancing AI event. Neither AMD, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE