Golf club maker Ping is not a typical supercomputing user. It's not making airplanes, looking for oil or investigating proteins. It is, instead, a company with 1000 employees that designs golf clubs.
But this mid-size company may represent a big part of the future of supercomputing.
Ping is using a Cray supercomputer built with Advanced Micro Devices Opteron chips to run simulations of golf club designs. And Eric Morales, staff engineer at Ping, said supercomputing has allowed the golf club maker to drastically reduce development time. Cray earlier this month announced Ping's use of its XD1 system, which can support 12 to 800 processing cores.
"It takes the development from weeks down to days, and it helps us get to market faster," said Morales. Simulations of product changes that once took a full day to run can now be processed in 20 minutes or less.
Morales believes many other midsize companies will turn to supercomputing.
Improvements in price to performance, due in part to systems using x86-based chips, are making supercomputing more accessible to businesses, according to vendors. The most recent Top500 supercomputer list shows large gains in the use of commodity chips.
Dave Turek, vice president of IBM Deep Computing, said demand for supercomputing is being driven by a number of forces, including the creation of new businesses that rely heavily on high-performance systems, such as digital animation and bioinformatics. Supercomputing "is also being driven by vast amounts of data that demand rapid analysis for real-time decision-making", he said.
Turek pointed in particular to RFID-enabled devices, which can be used to track product shipments, as capable of "generating huge amounts of data". While enterprise vendors are making capacity-on-demand systems available for processing large amounts of data, Turek doesn't believe those systems will account for more than 10 percent of the market.
Ed Turkel, manager of product marketing and high performance computing at Hewlett-Packard, said high-performance clusters, ranging from 32 to 64 nodes (a node is typically a two-processor system), are being increasingly adopted in industrial markets. Microsoft offers its own cluster product, which Turkel believes will help HP sell high-performance clusters to more commercial users.