Nvidia and Intel Introduce Cutting-Edge HPC Ventures, Enhancing AI Prowess at SC2023

Nvidia and Intel Introduce Cutting-Edge HPC Ventures, Enhancing AI Prowess at SC2023

Want to stay in the loop with the latest and greatest shifts in the AI-powered supercomputing field? Make a habit of checking out our daily and weekly newsletters full of the newest developments and exclusive insights.

The competition to push the boundaries of supercomputing speeds is heating up, with Nvidia and Intel at the forefront. These tech giants are neck and neck in enhancing the most powerful computers on the planet, and it’s all about boosting their AI capabilities.

Just recently, at Supercomputing 2023 in Denver, the tech community buzzed with excitement as the latest top 500 list of blazing-fast supercomputers was unveiled. Each one of these computing behemoths has a bit of Nvidia or Intel inside, and sometimes, they boast components from both. It’s a gathering that also sparks discussions on what’s next for these data-crunching monsters, exploring their future designs and applications.

Take Nvidia’s pride and joy, the JUPITER supercomputer. It’s under construction in Germany and is poised to be an AI powerhouse with an astonishing 90 exaflops of AI processing prowess, thanks entirely to its 24,000 Nvidia GH200 chips. At this event, Nvidia is also showcasing some serious tech wizardry with newfangled innovations aimed to propel AI processing to new heights, introducing the H200 and a jaw-dropping quad configuration for the Grace Hopper GH200 superchip.

Intel’s not just watching from the sidelines. It’s putting the pedal to the metal with the Aurora supercomputer over at the Department of Energy’s Argonne National Laboratory, where an eye-popping one trillion parameter language model is in the works. And that’s trillion with a ‘T.’ No typos here! Intel’s throwing open the curtains to give us all a glimpse into what’s coming in the AI and GPU acceleration arena, all while eyeballing Nvidia with a competitive glint.

The talk of the town, though, is Nvidia’s newfound gem, the Grace Hopper superchip. Having entered full production just this past May, these chips are at the heart of today’s top-tier supercomputers. Nvidia’s ushering in an era of AI supercomputers with Grace Hopper — and the JUPITER supercomputer is set to be a frontrunner. It’s going to be a game-changer in fields like weather forecasting, pharmaceuticals, and industrial design.

Nvidia’s not stopping there. The JUPITER will harness the might of the quad GH200 configuration, pulling off some serious computational gymnastics. Imagine a setup that fuses 288 Neoverse ARM cores and zooms along with 16 Petaflops of AI performance. It’s a head-turner. And all those GH200 chips are tied together with Nvidia’s next-gen Quantum-2 InfiniBand networking magic.

Intel’s got its own ace up its sleeve, boasting about its HPC and AI firepower at SC23. They’ve got a whole bunch of impressive hardware making waves, like the Data Center GPU Max series and the Habana Gaudi 2 accelerator. These tech treasures are the muscle behind the U.K.’s lightning-fast Dawn Phase 1 supercomputer. Not to be outshone, the Aurora will spawn one of the biggest large language models to date. Intel has been hard at work with Microsoft Deepspeed to make sure that this 1 trillion parameter AI goliath isn’t just for an elite few but can spark a revolution across various scientific domains.

If absorbing the latest tech news is your kind of thing, make sure to subscribe for a daily dose of hot-off-the-press updates right in your inbox. Who knows what fascinating developments tomorrow might bring in this accelerating world of supercomputing and AI!