SuperPi is one of the most popular benchmarks of today. Even though it’s very simple, and horribly single-threaded, there’s still a lot of prestige in being able to calculate pi to one million decimals the fastest. The current record is less than 7.5 seconds (7.437s to be exact, held by no_name) with the SuperPi application (it uses the Gauss-Legendre algorithm). FUGGER of XtremeSystems has now announced that SuperPi will make it’s way to GPUs now, through CUDA. The benchmark will be able to utilize the highly parallel design of modern GPUs, and should be ready within a month. We’ll keep you posted if anything comes up.
