I know it was only a demo (run by Nvidia) but it's exciting to read about the comparion of the power usage with the iPad. I think we are finally starting to see the emergence of some really power stingy chips. This will do a lot for making mobile computing even more in vogue.
That's a great point. Some of the things we should expect to see as the IoT era evolves will require extremely low power consumption devices that can remain in the field for many years with limited opportunities (at best) for replacement. Another things that I think makes this is an exciting time for the semiconductor industry.
What interests me are potential applications. We think of GPUs in the context of video adapters that deliver ever larger and more complex images to a screen, with 3D acceleration to handle animation, and results in high end games that can be hard to tell from a movie.
But GPUs have rather broader capabilities, and lots of things might do graphics processing to produce results that don't get displayed on a screen.
There are likely to be all manner of places chips like this might be used in IoT application.
I don't think I would classify this as just an NVIDIA advertisement. The two most impressive announcements this week at SIGGRAPH, other than the new specs, were the new Samsung Exynos 5 Octa processor with the Mali GPU and the NVIDIA Logan processor with the Kepler GPU. GPU's are a critical differentiator in mobile devices and will be important in IoT, as pointed out. GPUs can do much more with heterogeneous programming models like OpenCL, CUDA, and HSA.
Just think of Google Glass. It's cool, but not optimal. What if they could have minimized the CPU and memory, and focused on the GPU and connectivity? You would have an always connected solution. This does, however, assume a good connectivity that doesn't drain the battery. The point is that different IoT solutions will need to be optimized different requirements. The most critical will be connectivity, especially for mobile solutions. If they have consumer interfaces or interact with sensor data, GPUs are likely to be critical. So, these advances in GPUs are bringing computing to more platforms than anyone could have imagined 10 years ago.
Blog Doing Math in FPGAs Tom Burke 23 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...