The next generation operating system is getting a little smarter.
Operating systems manage programs and processes pretty simply, scheduling them in order of importance according to free cycles and running them on the CPU almost exclusively.
While this approach is more than acceptable, both NVIDIA and Microsoft reckon there's a world of performance to be unlocked if both the CPU and GPU can share the load.
They've worked together on a new tech called DirectX Compute, a new framework for Windows 7 that allows NVIDIA (and of course any other DirectX Compute compatible product such as ATI) GPUs to take over some processes from the CPU.
Workloads like video playback, on-the-fly video transcoding for portable devices and other highly parallel tasks can all be sent to the massively parallel GPU architecture, freeing up the CPU for other more traditional single-threaded tasks.
This is a great sign that they're finally thinking about how each piece of the computer should work together to solve problems, and to squeeze even more out of those GPUs can only be a good thing.
Head over to Tomshardware to read some more about the hybrid GPU/CPU workload sharing and DirectX Compute.
Issue: 107 | December, 2009