Facebook which is now touted as world’s largest social media platform is all set to make the design of its Artificial Intelligence server open source through Open Compute Project (OCP). Thus, by doing so, it is allowing other companies build similar systems.
Code named as Big Sur, the server helps Facebook in learning machine learning programs which help in image recognition, where a software program studies a photo or video to identify the objects in a frame. Such intelligence is also being applied to all kinds of large data sets, to spot things like email spam and credit card fraud.
Nowadays, companies like Facebook, Google, and Microsoft are pushing hard at Artificial intelligence, which helps in building smarter online services.
Facebook has also released some open source software’s serving and supporting Artificial intelligence (AI). For the first time it id now releasing the design of AI hardware.
As excepted, Big Sur depends heavily on GPU (graphical processing units), which are often more efficient than CPUs for machine learning tasks. Each AI processing server can have more than 8 GPUS that consume up to 300 units, and can be configured in number of ways via PCle.
According to Facebook’s press release on Big Sur, the said server doesn’t need any kind of critical cooling solutions, as a large air flow unit inside the server that presumably contains fans is deployed in the chassis. This helps in blowing cool air across the components and so cuts down the use of industrial cooling systems for keeping temperature of the whole system under control.
Big Sur is currently using GPUs from Nvidia, including its Tesla Accelerated Computing Platform. Its design availability through Facebook’s OCP will benefit all those companies which are interested in taking the help of AI software to study large data sets.