- Nvidia's (NASDAQ:NVDA) recently-launched Tesla M40 GPUs have been designed into Big Sur, a computing platform created by Facebook (FB) specifically for A.I./machine learning tasks, and which has been open-source for use/improvement by third parties.
- Big Sur contains eight GPU boards, but consumes just ~300W of power. Nvidia says it worked with Facebook to optimize Big Sur for delivering "maximum performance for machine learning workloads, including the training of large neural networks across multiple Tesla GPUs." Wired observes Facebook and others have found neural networks, which are modeled after the behavior of human neurons and improve their decision-making by taking in more data, are much more efficient when relying heavily on GPUs.
- Facebook, like Google, has been investing heavily in machine learning for some time. Use cases include translating content, detecting faces within Facebook photos/videos, and improving the company's news feed algorithm by analyzing text. Machine learning is also being used by Facebook's M assistant (still in test mode).
- The Facebook news comes a month after Nvidia unveiled the Jetson TX1, a module for drones/autonomous cars that runs on a Tegra processor and is declared to be "the first embedded computer designed to process deep neural networks." The company's Tesla GPUs are widely used in supercomputer/HPC deployments.