Sponsored Post: Analyst firm Gartner believes AI will remain one of the top workloads driving infrastructure decisions through to 2023 as more organizations push pilot projects into the production stage. But the company also notes that those initiatives require specific infrastructure resources that can grow and evolve alongside AI technology to ensure they are successful.
The simple truth is that AI is very compute heavy. Workloads engage in excessive number crunching that demands substantial processing power, memory muscle, and data storage resources, as well as fast, reliable, and low latency network connections to interlink those components. And while modern high performance computing (HPC) systems have stepped up to meet that demand to a certain extent, it’s equally true that many existing data centers haven’t yet come to the party.
Which leaves the traditional architectures in situ for the most part struggling to cope with the demands that AI is putting on them. But while organizations recognize the need for an upgrade, building out an IT environment customized to match the requirements of specific AI applications and workloads isn’t as easy as it sounds.
All sorts of factors need to be carefully considered — the balance of CPU and GPU resources to optimize utilization for example, as well as the precise composition of the racks, cabinets, servers, motherboards, memory modules, and disk drives (spinning and SSD). And for all of those components to work efficiently in tandem, they also need a lot of bandwidth at both system and network level to smooth the transmission of the large volumes of data that AI models tend to create.
You can learn more about these challenges by listening into this webinar — “Meeting the Bandwidth Demands of Next-Gen HPC and AI System Architectures” — at 9 am PST on November 9, 2022.
Moderated by The Register’s Nicole Hemsoth, the panel features a stellar cast of knowledgeable speakers. They include Vladimir Stojanovic, co-founder and chief architect at Ayar Labs; Simon Hammond, program manager for the National Nuclear Security Administration; Christopher Long, chief architect at Liqid; and Alex Ishii, a distinguished architect at Nvidia.
Together, they’ll explore the innovative system designs and solutions they are building to address the AI/HPC infrastructure problem and help unlock the huge potential that AI presents to research, science, and business initiatives.
You can register to watch the “Meeting the Bandwidth Demands of Next-Gen HPC & AI System Architectures” webinar by clicking here.
Sponsored by Ayar Labs.
How – And When – Optical I/O Will Make Disaggregated Systems Better
As many of you know from reading The Next Platform, we are firm believers that eventually we will get disaggregated and composable systems that drive up the sharing of hardware resource across many workloads and therefore drive down the cost of hardware to support workloads. This would have always been …
HPE Taps Ayar Labs To Add Silicon Photonics To Slingshot Interconnect
There are so many barriers to capacity increases in high performance computing systems that it would take a long time to rattle them off. But one of the most important ones, which we have covered in-depth with many experts in recent years, is the day that electrical signaling can no …
Nvidia Shows What Optically Linked GPU Systems Might Look Like
We have been talking about silicon photonics so long that we are, probably like many of you, frustrated that it already is not ubiquitous. But the good news is that advances in electrical signaling, after hitting a wall a decade ago when the talk of practical silicon photonics interconnects first …
Be the first to comment