Sparxity announced a multi-million-dollar deployment of its sparse-native computing system at the Jülich Supercomputing Centre. The company, formerly SpiNNcloud, builds systems that compute only what matters. The installation is named JUBIS.
System performance
The system is installed at JSC, one of Europe’s largest supercomputing research centers. It is designed to run 10 times faster for sparse solver execution. It also targets 18 times lower energy use for event-based machine learning and 20 times shorter optimization solve times.
Company rebrand
Sparxity was formerly known as SpiNNcloud. It builds full-stack systems from silicon to software. Its systems are deployed in rack- and supercomputer-scale installations across Europe and the US.
How the technology works
The system uses an event-based approach that computes only when something happens. It reduces the amount of computation and data movement required. Current systems serve HPC and AI, and the next generation focuses on sparsity for AI inference.
The installation is supported by the European Regional Development Fund and the State of North Rhine-Westphalia. The system is intended to become available for wider HPC and AI applications. JSC’s adoption provides another proof point for sparse-native computing in real infrastructure.