For Facebook, artificial intelligence is absolutely the key to its future. The social networking giant depends on AI to deliver more relevant news feed posts, photos, and videos, as well as power speech and text translation. Underlying all that work is a series of powerful servers that Facebook designed and built specifically for the task. Since 2015, it has been using a GPU server known as Big Sur.
Now, Facebook is moving on to the next-generation GPU server, known as Big Basin, which it announced today in conjunction with the Open Compute Project. The new servers, the design of which the company is open-sourcing, will allow for training “machine learning models that are 30% larger because of the availability of greater arithmetic throughput and a memory size increase from 12 GB to 16GB,” it says. That potentially means more powerful and faster AI operations across Facebook.