by Andy Walsh

Who knew that one of life’s simple pleasures – ice cream – could be so complicated?

Scientists at the University of Edinburgh, in Scotland, are using a supercomputer to research – at the molecular level – how to improve the texture and shelf life of the world’s favorite frozen concoction.

Dr. Alan Gray, one of the lead scientists in the program, says that ice cream is far more complex than most people realize. Its constituent substances interact in a large variety of ways over time (think of all the ingredients in my favorite flavor, Ben & Jerry’s Chunky Monkey). It turns out that the best way to understand those interactions is to simulate them, but the simulations would take years to model on personal computers.

So the team at the Edinburgh Parallel Computing Centre (EPCC) turned to supercomputers to speed up the simulation work. The team conducted early research on a huge Cray supercomputer with over 200,000 CPU cores in 200 cabinets. But they realized that smaller, GPU-accelerated systems offered the same level of capability for their needs.

The EPCC team turned to a 10-cabinet Cray XK6 hybrid system, which integrated 936 NVIDIA Tesla GPUs with standard x86 CPUs to turbocharge the application. The team soon found that performance on the Tesla-accelerated XK6 was two and a half times faster than using CPUs alone.

In the brief video above, Dr. Gray explains that this research is applicable to large number of other soft materials, including motor oil, paint, catsup, mayonnaise, yogurt, and cleansers like soap and shampoo.

Their research reflects a growing trend in high-performance computing to use simulation to improve the quality of thousands of products we use every day, from golf clubs to medicines to automobile engines.

Of course, ice cream is the coolest example.

For more information, read the case study.