Artificial intelligence is driving a major sea change in the high-performance computing world.

At the Supercomputing 2023 event last week, experts from throughout the HPC ecosystem gathered in Denver, Colorado, to assess the state of their industry. What they found was a technology on the brink of a transformative era, one that will be driven by the integration of AI in traditional computing, combined with advancements in semiconductor and cloud architectures.

“AI is the gift that’s just giving more and more energy — and hype — to this market of technical people,” said theCUBE’s John Furrier, industry analyst for theCUBE, SiliconANGLE Media’s livestreaming studio, during a live analysis at the event. “But that’s the theme of the show so far. The impact of AI to high-performance computing and the intersection of the semiconductor players and the cloud is happening. Everything in between is going to be up for grabs for innovation.” (* Disclosure below.)

Furrier was joined in the segment by fellow analysts Savannah Peterson and Lisa Martin.

Here is theCUBE’s analyst discussion during day 2 at SC23:

Here are three key insights you might have missed during the event:

1. Growing demand for AI is creating new opportunities for HPC in the research community.

The application of AI in HPC is opening new vistas for research at major U.S. technology research facilities. At the Ohio Supercomputer Center, under the auspices of the Ohio Department of Higher Education, researchers in partnership with Dell Technologies Inc. are leveraging AI and HPC to develop innovative approaches to weather modeling.

“In the past, you would run a simulation and that model will tell you, ‘OK, here. This is where we think the hurricane’s going to hit,’ for one example,” said Armando Acosta, director of HPC product management at Dell, in an interview on theCUBE. “But can you imagine now, you take that result out of that simulation and you plug it into a neural network. You run training and that insight combined with that model now gives you two different perspectives so you can actually get a better insight without each individual one — that’s the beauty of AI.”

Here’s theCUBE’s complete video interview with Armando Acosta, who was joined by Alan Chalker, program director of the Ohio Supercomputer Center at the Ohio Department of Higher Education:

Dell’s technology has also found a home at the University of Cambridge where researchers are using the company’s liquid-cooled Intel Ponte Vecchio variant system as a converged AI and simulation platform. It is now the fastest AI system in the United Kingdom, according to Paul Calleja, director of research computing at the University of Cambridge, in conversation with theCUBE at SC23.

“[Dell’s PowerEdge XE9640] is designed to kind of be a converged AI and simulation platform,” Calleja said. “We’ve been working with Dell for a long time to get to that stage in co-design in the box. Working inside in the software environment … there’s a lot of applications already that we can put on that system that are going to make some real impact.”

Here’s theCUBE’s complete video interview with Paul Calleja, who was joined by Andrea Booker, director of product management for PowerEdge Servers at Dell:

One issue that could hamper growth of AI and HPC is accessibility to hardware. The Nvidia Tensor Core H100 GPU is a popular integrated hardware and software solution for AI, but it is also in high demand, as viewers learned in theCUBE’s interview with Brian Beeler, editor of StorageReview.com.

“In terms of the where question, can I get access to the hardware?” Beeler said. “We were told just a week and a half ago that the wait for H100 systems, if you place your order today, is somewhere around 40 weeks. Not everyone needs H100, and Nvidia says, ‘Well look, we’ve got this shiny L40S; it’s available now.’ It’s opening up the market. I think so too for AMD, Intel and others with accelerators to come in with perhaps lower cost options that may not be quite so powerful.”

Here’s theCUBE’s complete video interview with Brian Beeler:

2. Broadcom is playing a central role in shaping HPC technology.

Broadcom Inc. has been a large-scale supplier of enterprise networking and connectivity hardware for decades, so it is no surprise that the company would be a prime participant in technology for the HPC world. Rapid advances in AI and machine learning are driving Broadcom’s business as well.

“At Broadcom, we have been doing ethernet networking for decades now,” said Hemal Shah, distinguished engineer and system, software and standards architect at Broadcom, during an appearance on theCUBE. “AI/ML is pushing the envelope, bringing really large-scale clustering, asking for high bandwidth. And what we have is both in terms of our [network interface cards] and switches, we have our solution today. We are innovating with other things around ultra ethernet, where we are putting more intelligence in network infrastructure in the NIC to make the AI/ML solution from a networking standpoint the best.”

Here’s theCUBE’s complete video interview with Hemal Shah, who was joined by James Wynia, director of product management for networking, Enterprise Solutions Group, at Dell:

Increased enterprise reliance on large language models is creating demand for new tools to drive and measure performance. One tool in this area is MLPerf, an open-source benchmark suite for measuring how fast systems can run models in a range of scenarios, a central focus for Broadcom.

“My job is I’m trying to look at what the workloads look like,” said Kimberly Leyenaar (pictured, left), principal storage performance architect at Broadcom, in a discussion with theCUBE. “We’re working a lot with MLPerf trying to drive these workloads … we have the tools to kind of analyze what that workload looks like from the switch and from the network and from the storage.”

Here’s theCUBE’s complete video interview with Kimberly Leyenaar, who was joined by Steen Graham (right), co-founder and chief executive officer of Scalers AI Inc., an enterprise AI software company:

One of the key areas of focus for both Broadcom and Dell has been formulating industry standards in optimizing fabrics that can support AI platforms. Broadcom has been an active proponent of the peripheral component interconnect express, or PCIe, interface standard for connecting high-speed components and has decided to take an open approach, according to Jas Tremblay, general manager for the data center solutions group at Broadcom, in an appearance on theCUBE.

“We believe the right answer is PCIe with Compute Express Link extensions,” Tremblay said. “We’ve decided this year to be completely open, non-NDA with a roadmap. It’s sharing the fabric roadmap with the entire industry, so they know what’s coming.”

Here’s theCUBE’s complete video interview with Jas Tremblay, who was joined by Ihab Tarazi, senior vice president and chief technology officer of Dell:

3. HPC and AI are driving changes in the storage industry.

The emergence of HPC and AI as key technologies in enterprise computing has positioned networking, data processing and server storage as important resources in the IT picture. As noted by several storage industry experts interviewed by theCUBE during SC23, the integration of AI into HPC is opening new customer conversations in the storage field.

“You can see where there’s a seismic delta between HPC systems that are fast but boutique and fragile and enterprise storage, which is robust and available and online, but not performing enough to meet the needs of accelerated computing,” said Vaughn Stewart, vice president of systems engineering at Vast Data Inc., in an interview on theCUBE. “You sit down with the founders of a company who are like, ‘We’ve identified this.’ First is to solve the performance cost scale challenge. Next is to go and start to fold in all these data services, data processing platforms to give a more richer set of leveraging your data, reducing time windows, reducing moving data around, reducing cost.”

Here’s theCUBE’s complete video interview with Vaughn Stewart:

The need for speed in data processing has led one major vendor to significantly boost its input/output operations per second, or IOPS.

“We announced the fastest system in the industry for storage,” said David Wohlford, worldwide senior product marketing manager for IBM Storage for AI and cloud scale, during his appearance on theCUBE. “We already are the fastest storage, but we now doubled it … we can now bring 277 gigabytes per second of our storage. You talked about having 13 million IOPS in a rack; we could put 130 million IOPS in a rack now, I mean, faster than any system out there.”

Here’s theCUBE’s complete video interview with David Wohlford:

IBM’s activity in the storage field has also led to the introduction of new solutions to access massive amounts of data and deliver that to applications, regardless of where these reside in the enterprise. Sycomp, A Technology Company Inc., has been working with IBM and Google LLC to provide a performant parallel file system that can dynamically access data in Google Cloud Storage and deliver it with low latency.

“With IBM Cloud and Google Cloud … they’ve built the networks, they’ve built the infrastructure,” said John Zawistowski, global systems solutions executive at Sycomp, in conversation with theCUBE. “The challenge is data … how do we get it to the application regardless of where it is? That’s what we’ve worked on with Google and IBM Cloud to bring them a solution that allows that data highway between anywhere and anyplace.”

Here’s theCUBE’s complete video interview with John Zawistowski, who was joined by Hugo Saleh, head of HPC at Google, and Leon Platts, vice president of IBM Cloud high-performance computing:

To watch more of theCUBE’s coverage of the SC23 event, here’s our complete event video playlist:

(* Disclosure: TheCUBE is a paid media partner for SC23. Neither Dell Technologies Inc., the main sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE