Case Study: Electric Sheep
Exploiting the full potential of multi-core processors
Download the full document as a PDF
Electric Sheep is a global computer-generated art project, harnessing the power of the Internet to render intricate fractal images which are then composed into animated film sequences. With today’s singlecore CPUs, a single frame from such a rendering takes seconds to compute.
Using the RapidMind platform, these images are generated at interactive rates of over 10 frames per second. With RapidMind, Electric Sheep can take advantage of state-of-the-art multi-core and stream processors such as GPUs to achieve this breakthrough performance.
What does this mean for you? With the RapidMind Platform you can more quickly and easily build higher-performing applications that for the first time exploit the full capabilities of multi-core processors, including the Cell Broadband Engine™ (Cell/B.E.) from IBM and graphics processor units (GPUs) from AMD® and NVIDIA®.
Electric Sheep
In homage to Philip K. Dick’s novel Do Androids Dream of Electric Sheep (later popularized in the 1982 major motion picture, Blade Runner) Electric Sheep is a screen saver created from fractal images (called sheep).
Flame fractals are rendered as JPEG images by distributed client software, which is then transferred across the Internet to the Electric Sheep server. The server stitches these images together in ever changing fractal sequences as compressed 640x360 MPEG-2 video.
Generating one fractal usually takes multiple seconds, and is not generally rendered in real-time.
In this demonstration
The fractal generation computation of Electric Sheep was implemented on the RapidMind platform, and run on an NVIDIA GeForce® 8800 GPU. For comparison, the standard Electric Sheep program was run on an Intel® Duo 6700 Dual-core CPU.
- Render Speed with Electric Sheep: 0.16 frames/second
- Render Speed with RapidMind: 10.0 frames/second
60x performance increase with RapidMind
The results demonstrate how RapidMind performance scales as additional processing resources are available. Computations once confined to offline processing can now be performed realtime on a single computer with a GPU.
Watch the Video:
Quicktime: Large (15.1 MB)