Cray XK6 Supercomputer Used to Analyze Ice Cream

Researchers in Scotland are using the massive system to find out how the various ingredients in ice cream interact and, in turn, improve other common products.

Researchers at the University of Edinburgh are using a Cray supercomputer to make ice cream taste better and last longer.

The Scotland university is leveraging a 10-cabinet Cray XK6 supercomputer€”armed with a combination of standard x86-based Opteron chips from Advanced Micro Devices and 936 Tesla graphics technology from Nvidia€”to simulate ice cream with the aim of getting a better understanding of how the various ingredients that go into making ice cream interact with each other.

With that knowledge in hand, the researchers with the Edinburgh Parallel Computing Centre (EPCC) can improve the taste of ice cream and give it a longer shelf life, according to officials with Nvidia. In addition, the knowledge gained from the research into ice cream can be applied to a host of other products that people use every day, from paint, mayonnaise and engine oil to yogurt, soap and shampoo, according to Alan Gray, one of the lead scientists in the program at EPCC.

While turning the power of a supercomputer like the XK6 to a subject like ice cream may sound a bit frivolous, Gray and Cray official Jay Gould said that like other products commonly in use today, the underlying substances and the way they interact are complex, and a powerful system is needed to run the necessary computations.

"If you zoom in on the substances and if you're able to view them microscopically, you would see they're really made up of quite complicated structures of different materials," EPCC's Gray said in a video about the project.

Cray's Gould said that using anything less powerful than a supercomputer would take far too much time.

"To get down to the details of doing simulations and modeling the world as we know it€”to better understand it€”you've got to have a lot of compute power €¦ not just a laptop," Gould said in the video. "If you used a laptop to generate some of these computations, it might take five lifetimes. We don't have time to do that."

Cray and Nvidia are touting the Edinburgh ice cream project not only as a way to show how supercomputers can play a significant role in better understanding everyday life, but also the performance advantages of using hybrid systems, which leverage both traditional CPUs, like those from AMD and Intel, as well as GPUs as coprocessors.

According to a post on Nvidia's blog by Andy Walsh, director of marketing for the company's Tesla business, the EPCC team initially conducted research using a Cray supercomputer powered by more than 200,000 CPU cores in 200 cabinets. They then switched to the XK6, realizing that "smaller, GPU-accelerated systems offered the same level of capacity for their needs."

The use of such coprocessors and accelerators is gaining traction in the supercomputing world. The organization that creates the Top500 list of the world's fastest supercomputers found that in the latest list, released in June, 58 of the systems used coprocessors and accelerators, up from 39 supercomputers six months earlier. Fifty-three of those 58 supercomputers use Nvidia chips, while two use Cell processors, another two use ATI Radeon GPUs from AMD and one system is running Intel's new Many Integrated Cores (MIC) coprocessor technology.

Nvidia chips are found in three of the top 10 fastest supercomputers, according to the Top 500 list.

Nvidia's Walsh said the EPCC ice cream project is part of a larger trend in the high-performance computing field of using supercomputers and simulations to "improve the quality of thousands of products we use every day, from golf clubs to medicines to automobile engines." Cray's Gould said projects like the ice cream research by the EPCC will have far-reaching impacts.

"If we can control the way phase separation occurs and those kinds of things for ice cream, we can leverage it into all kinds of other, more important technologies as well," he said.