In what’s turning into an annual custom for the @HPCpodcast, we current “Dwell from Nvidia GTC 2025,” masking highlights from the Nvidia extravaganza with an AI-everywhere theme.
We overview the convention, discussing all the pieces from the brand new AI compute trade panorama, together with the latest previous (Hopper), the current (Blackwell) and the long run (Rubin and Feynman) — i.e., the Nvidia GPU roadmap — plus silicon photonics for cluster interconnect cloth (which we imagine was the star of the present), the complexity of inference, low-end programs, energy and cooling (did we hear 600 KW per rack?), software program (together with cluster-level AI-workload-focused open-sourced Dynamo “OS for AI”) and Jensen Huang’s up to date definition of storage (the semantic sort), it’s not nearly retrieval any extra.
You’ll find our podcasts at insideHPC’s @HPCpodcast page, on Twitter, on the OrionX.net blog page, on iTunes and Google. Right here’s the OrionX.net podcast page, and the RSS feed. We’re additionally accessible on Spotify and iTunes.