GPUs offer analytics boost

SAN JOSE--General-purpose computing on graphics processing units (GPGPUs) are increasingly used in high-performance computing (HPC) by certain industries for data analytics but widespread enterprise adoption will take some time due to lack of applications and support by independent software vendors, industry players note.

Cyrus Golkar, executive vice president of business development at in-database analytics software company Fuzzy Logix, said GPGPU is suitable for data analytics because both technologies are based on numbers and graphics processors are "really good" at crunching numbers.

Elaborating, he said during an interview here that the company helped a stock exchange develop a proof-of-concept which calculated a billion scenarios in 13 milliseconds using only one medium-size server and GPU-based analytics. This, he added, would not have been possible even with more than 10 times the number of CPU-only servers.

Despite the advantages on offer, Golkar noted that enterprise adoption of GPGPU is still at the early stage. He believes uptake will pick up once people see what they can do with the technology, though.

Sumit Gupta, senior director for Tesla GPU computing at Nvidia, said in a separate interview during the company's GPU Technology Conference held here this week that it is in the high-performance computing arena that has adopted GPGPU for analytics much faster because it needed to address bigger pain points.

For enterprises to utilize GPU computing for analytics, the executive noted that it will "be all about applications". Independent software vendors, for one, will have to spend a longer time adopting the technology since they need to ensure their products are "100 percent solid" and able to run on the new computing platform, Gupta said.

Golkar added that while it is "easy" to include GPUs into computing systems, it will take certain expertise to program and build applications to leverage the computational capabilities on offer. Professionals such as quantitative analysts with their mathematical backgrounds are examples of those who are sought after to tap GPGPU.

Fuggy Logix, for one, is planning to set up its research and development center in China in the near future to leverage the country's strong talent pool in mathematics, he revealed.

Liau Yun Qing of ZDNet Asia reported from Nvidia's GPU Technology Conference in San Jose, United States.