You may have heard of Vela, IBM’s AI supercomputer. Vela meets customers with varying compute needs for GenAI model training right where they are.
Born in the IBM Cloud in 2022, Vela was designed keeping in mind the variable compute demands across the different stages of the AI lifecycle. Many IBM’s internal AI technologies have been trained and prototyped on this supercomputer.
While Vela continues to provide extensive support for AI workflows, IBM is stepping up the game. It is expanding Vela’s capabilities. Last year, the company started development of a more powerful version of Vela, named Blue Vela, that will support IBM Research’s AI training jobs.
It is said that Blue Vela is threefold faster, and packs significant compute capacities for bigger model training jobs. That opens door for IBM to leverage it for rapid development of its most complex models. Unlike Vela, the Blue Vela cluster is stationed in IBM’s on-premise Research datacenter.
On Gestalt IT Rundown this past week, hosts Stephen Foskett and Tom Hollingsworth discusses the architecture and performance of Blue Vela, telling the audience what to expect. Take a listen, and for more stories like this one, sign up for the Gestalt IT newsletter today.
