Deep learning is the capacity for a computer to recognize specific representations (images, texts, videos, sounds) after being shown many examples of these representations. For example, after being introduced to thousands of cat pictures, the program “discovers” by itself what are the specific features of a cat and can then distinguish a cat from a dog or any other picture.
This learning technology, based on artificial neural networks, has revolutionized Artificial Intelligence (AI) in the last five years. So much so that today, hundreds of millions of people rely on services powered by deep learning for speech or face recognition, real-time voice translation or video discovery. It is used for example by Siri, Cortana and Google Now.
The Atos and Bull teams are actively working, together with technological partner NVIDIA®, to provide the Deep Learning expertise, the hardware and software solutions, and the services you need throughout the development stage and the production stage of your Deep Learning project.
Bull sequana – the open supercomputer that is ready to take on the major challenges of the 21st century – supports different types of compute blades, including the GPU-accelerated Bull sequana X1115 blade. With this compute blade, Bull sequana leverages the NVIDIA Pascal architecture, with Tesla® P100 and NVLink, to deliver up to 50x performance boost for applications and drive new possibilities in Deep Learning applications.
Thanks to an agreement between Atos and NVIDIA, the NVIDIA DGX-1 server is also part of the Bull portfolio, as a complement to Bull sequana for Deep Learning needs. Together they offer a complete solution to learn, see, and simulate our world. The Bull sequana X1115 topology is similar to the NVIDIA DGX-1. They share the same NVIDIA Tesla P100 accelerators, NVIDIA NVLink 2.0, and other NVIDIA technologies. This makes it easy to develop / port / test / optimize your project on NVIDIA DGX-1 and then have a seamless transition to production on Bull sequana at extra-large scale.