THE UNITED STATES Defense Advanced Research Projects Agency (DARPA) is looking forward to using graphics processing units (GPUs) to help process vast amounts of data for government and military applications, and Nvidia is, of course, rather excited about the prospect.
In a company blog post, Nvidia highlighted a DARPA presentation at the GPU Technology Conference this week in which DARPA project manager Chris White told attendees of the agency's keen interest in GPU computing. He said, "The GPU has enormous potential for DARPA in solving analytics problems."
White revealed that DARPA has more than a dozen projects that target data analysis, but said that only two of those presently are using GPUs. He added that one of them, the XDATA cloud project, applies GPU computing technology to process and analyse large, imperfect and incomplete data sets, and admitted that DARPA is actively looking for additional GPU programming talent. He said, "We don't have many people who can program for it."
White indicated that the analysis of big data sets is a growing problem for DARPA's military customers, and said that XDATA is one of its approaches for developing more efficient software to address those needs.
He said he attended a recent White House meeting with government and military leaders where the issue came up. "They all said the same thing," he said. "The data are changing faster than they can adjust."
DARPA thinks using GPUs to process data analytics can help, but White said that GPU compute programming has proven to be a challenge for the agency. "There are many analytics problems that need solutions, but right now we can’t do it well or quickly enough with the GPU," he said.
Naturally Nvidia is keen to attract and help train skilled developers to take up GPU programming, both at DARPA and in scientific research establishments.
It might first have to lure them away from developing video games, however. µ