New in our CIO Insight Exchange
section, an article from the pages of the recent print edition of eWEEK
which brings us up-to-date with current developments in supercomputing. Jeffrey Burt
writes about advances in performance and energy efficiency achieved through the use of "the graphics processing unit (GPU)":
Sumit Gupta, director of marketing for Nvidia’s Tesla products, told eWEEK that GPUs hold several advantages over traditional chips—in particular, they’re much more energy efficient and can handle parallel processingworkloads. They’re also smaller and more efficient: Whereas AMD and Intel are offering as many as 16 cores per chip, Nvidia’s GPUs can hold as many as 500 cores.
The challenge, Gupta said, is to help organizations enable their workloads—the bulk of which are optimized for sequential processing—to run in parallel processing environments. Most applications run faster in parallel environments, but businesses and IT professionals are more comfortable with sequential processing. Given the performance enhancements and energy efficiency of parallel computing, though, that will change.
“The world is parallel,” Gupta said. “There’s no going back.”
A fascinating look at a part of the tech world we don't often hear about, but one which is likely to impact the way more mainstream computer functionality will be delivered in the future.
Graphics accelerators grow presence in supercomputing
Jeffrey Burt | 26 November 26, 2011