Facebook recently announced open sourcing hardware design for its custom designed Open Rack compatible hardware. Attributing advances in Machine Learning and Artificial Intelligence to richer data sets and more powerful GPU-based systems, Facebook is unveiling its next generation systems code-named “Big Sur”, after the synonymous location in California.
Big Sur is powered by eight Nvidia Tesla M40 GPU’s, each consuming up to 300 Watts. The design can support a wide range of PCI-e cards other than Nvidia Tesla, according to Facebook. Using off the shelf components, this hardware design can be used in standard Open Compute compliant data centers.
Facebook has previously open-sourced deep-learning modules for Torch, a scientific computing framework with wide support for machine learning algorithms. Torch modules alongside with open-sourcing Big Sur designs can benefit greatly academics and startups that want to build AI systems lacking Facebook’s resources.
Google last month open sourced its own machine learning library, TensorFlow which was received with mixed feelings by the data community. Microsoft also recently open sourced DMTK, a Distributed Machine Learning Toolkit of its own. Interest in deep learning has been exponentially increasing throughout the past years both on the software and the hardware side. Open sourcing frameworks and hardware designs means better and faster access to the large community of practitioners out there which can only be welcomed.
