Cornell University’s Center for Advanced Computing (CAC) announced a new initiative to test GPU-optimized MATLAB for use on various research projects. They’ve partnered up with Dell, NVIDIA, and Mathworks to see what GPUs bring to the table in terms of the university’s research.

Cornell cites a couple of interesting data points in their press release (here link1). The first is a 15x speed-up (well, 14.7 to be tediously accurate) in processing images used to diagnose cancer cells. Pre-GPU, it took 86.9 seconds to process a single image. Post-GPU, that time plummets to 5.9 seconds. The benefit is obvious – this increases the theoretical maximum number of images they can process from 994 per day to 14,644.

But that’s not all… (see below…)

I imagine that there are plenty of you out there saying, “Sure, I guess it’s a good thing that cancer research gets a boost from GPUs, but all I really care about is how they can be used to identify bird species from the noises they make in flight.”

Good news on that front as well: the new MATLAB with CUDA optimization has yielded a 12x speed-up on Dynamic Time Warping (DTW). This is the most computationally intense part of figuring out what sort of bird is making which horrible squawking sound.

Using the noises they make in flight is faster and “arguably” more accurate than actually eyeballing the birds as they fly overhead carpet-bombing newly washed cars. Networks of sound sensors will be able to track bird migration patterns, which is useful information if you’re looking to find them and get a little payback….or if you’re studying them too, I guess.