What Will GPU-On-The-CPU Mean for Analytics?
Posted by Bob Warfield on January 21, 2011
This week’s InfoBoom-sponsored post is all about Intel’s announcement that it would be shipping chips that include an integrated graphics processor (GPU) on the same chip as the CPU under their so-called Sandybridge architecture. A lot of folks probably ignored the announcement thinking it meant better video games for their kids and perhaps their laptops, but there is a more interesting way to look at the impact of such architectures for business.
GPU’s are not just for graphics, as it turns out. They’re extremely powerful supercomputers in their own right, with vector processing and other capabilities that are well on par with the Cray supercomputers that ruled during my college computer science days. The Cray X-MP ran 941 Megaflops back in the day and was considered strategic weapons grade computing. The latest Nvidia CUDA GPU’s weigh in at 1 Teraflop or better, so are fully capable of keeping up with the old Cray. What happens when that kind of power is available on every CPU? A whole lot of power, and a whole lot of change.
Check out my post over on InfoBoom to find out!