News Hub

NASA taps Nvidia GPUs to predict solar flares with cutting-edge deep learning

Written by Wed 2 Sep 2020

Space agency using Nvidia Quadro RTX-powered HP Z8 data science workstations to map the flow of the sun’s surface and predict solar flares

NASA researchers have developed new deep learning techniques powered by Nvidia GPUs that can understand what’s happening beneath the sun’s surface and predict earth-damaging solar flares.

The intense heat created by our nearest star creates a boiling reaction which makes its surface bubbly. These bubbles (or granules) are visible when magnified through telescopic images and offer an indication of what’s happening beneath the sun’s outer layer.

With its new state of the art AI setup, the US space agency can measure the flows of plasma in the sun’s atmosphere to predict its evolution, including events like solar flares which can bring down power grids, satellite communication systems, and even put astronauts lives at risk.

Interstellar innovation

Advanced imaging techniques are required to track the movement of the sun’s granular surface, which NASA researchers developed using data science and GPU computing with Nvidia Quadro RTX-powered HP Z8 workstations.

NASA crafted custom AI algorithms with a deep learning neural network which observes granules using images from the Solar Dynamics Observatory and then learns how to reconstruct their motions.

Nvidia GPUs were needed to train the neural networks as ordinary CPU power couldn’t handle the number of iterations or the amount of preprocessing required to develop robust deep learning models.

Before the researchers switched from a 72 CPU-core compute node to an Nvidia Quadro RTX 8000 GPU, it took an hour to complete one pass with the training data. The space agency’s new set up now processes one training round in about three minutes.

“This incredible speedup enables us to try out different ways to train the models and make ‘stress tests,’ like preprocessing images at different resolutions or introducing synthetic errors to better emulate imperfections in the telescopes,” said Raphael Attie, solar astronomer at NASA’s Goddard Space Flight Center. “That kind of accelerated workflow completely changed the scope of what we can afford to explore, and it allows us to be much more daring and creative.”

Written by Wed 2 Sep 2020


deep learning gpu hardware nvidia space
Send us a correction Send us a news tip