Mike Vizard has a piece in CTOEdge that shows just how easily you can miss the mark on a press release. He starts off with the recent comments from NVidia’s Sumit Gupta about NVidia making a play for HPC with new Tesla and Project Denver designs.
What’s about to change, however, is that access to graphical processing units (GPUs) is about to get a whole lot less expensive thanks to cloud computing services. What that means, says Sumit Gupta, who heads up product management and marketing for Nvidia’s Tesla products, is that mainstream enterprise software, such as business intelligence applications, will be embedding visually-oriented analytics that will be processed in the cloud.
Then he goes on about how in the future you’ll be able to use the cloud of video encoding and gaming.
The real power behind NVidia’s argument is in the growing mountains of data. Data that used to be easily visualized with a simple Bar Chart or Line Graph has grown beyond the ability of simple Windows Graphics to render with any usefulness. You now need more advanced visualization methods like Treemaps and Heatmaps, not to mention Edge Bundles and network graphs. These algorithms generally require hardware acceleration (if not require, then benefit greatly from) due to the complex nature of the computations involved (Edge bundles) or the difficulty in rendering (heatmaps).
Just imagine if Google or Tableau extended their charting ability to include iRay-style visuals (resolution-varying streamed rendering) of massive datasets.
