Slide View : Parallel Programming :: Fall 2019 (original) (raw)

Why Parallelism? Why Efficiency?

Previous|Next--- Slide 46 of 52

Back to Lecture Thumbnails

barracuda

ORNL had already started adding GPUs to its supercomputers with Titan, Summit's predecessor. Titan had almost 19,000 GPUs, with a peak performance of 27 petaflops. Summit has fewer CPUs (but more cores per CPU) and more GPUs than Titan and has a peak performance of 200 petaflops, demonstrating the role of parallel computing in increasing overall performance.


Copyright 2019 Stanford University