OpenCL Diagram showing integration between CPU and GPU
The Khronos Group announced the ratification and public release of the OpenCL spec earlier this week. OpenCL is a new industry-standard framework that will allow the powerful but specialized chips on graphics cards to be used for crunching numbers, just like the general purpose CPU in a computer.
That’s a lot of geek-speak to say that a lot of time-consuming multimedia tasks are about to get a lot faster.
Today’s graphics cards are capable of some serious mathematics, but they’ve always been restrained to using it to push pixels live to the screen. Essentially, a graphics card contains thousands of cores, all of which work on a small piece of the screen image in parallel, or at the same time. A general purpose CPU, on the other hand, typically has fewer cores (today’s most powerful desktops have 4 or 8), so they must process operations serially, or one after another.
With technology like OpenCL, a GPU’s raw processing power can be harnessed for some general computing tasks which can be broken down into independent steps which can run at the same time, like complex simulations or video compression.
The next version of Apple’s Mac OS X 10.6, expected for release in 2009, will feature this technology. The upshot for media producers like myself is that sometime in the next year, we will spend less time rendering and compressing our work, freeing up more time to perfect the design itself.
