Introduction
General purpose graphics processing units (GPGPUs) have had a huge impact in everything from professional and home applications to video processing and even physics simulations. No GPU parallel computing architecture has been more in the spotlight than NVIDIA’s CUDA either. CUDA technology is being used in many different types of applications including video and audio encoding, oil and gas exploration, product design, medical imaging, and scientific research.
CUDA-enabled apps give CPUs a break during video transcoding
General purpose graphics processing units (GPGPUs) have had a huge impact in everything from professional and home applications to video processing and even physics simulations. No GPU parallel computing architecture has been more in the spotlight than NVIDIA’s CUDA either. This Instruction Set Architecture (ISA) uses a parallel compute engine in NVIDIA’s graphics processing units to solve many complex computational problems significantly faster than a CPU. CUDA performs two major functions that consumers should be aware of – it helps reduce or match CPU usage by engaging the GPU’s stream processors and it can accelerate any computing process that CUDA is enabled.
Another important factor to note is that CUDA is supported by many lower-end NVIDIA GPUs including the GeForce 8400GS, 8500GT, 8800GTS, 8800GTX, 8800 Ultra, and all the way up to the GTX 295 (complete list of compatible video cards available here). Professional GPUs like Tesla-series video cards or GeForce and Quadro-series mobile GPUs are also compatible with CUDA technology. This means that consumers on a budget or those with older-generation NVIDIA GPUs can take full advantage of graphics acceleration CUDA enables in many applications.
CUDA technology is being used in many different types of applications including video and audio encoding, oil and gas exploration, product design, medical imaging, and scientific research. So what CPU function has CUDA been most utilized by software developers recently? The answer is consumer-level video transcoding applications.
(Graphics provided by Loilo, ArcSoft, MotionDSP, Elemental, and Cyberlink)
The ideal transcoder should be simple to use, reduce or match the computing load placed the CPU, produce a high quality reproduction of the input, and complete the transcode as fast as possible. Typical CPU-based transcoding usually makes the primary processor do all the work. CUDA-enabled GPUs allow the graphics card to process all the computations and leaves only very specific work for the CPU.
With that in mind, we selected five CUDA-enabled transcoding applications that use this architecture is similar, but in some instances, very different ways. We chose to test MotionDSP’s vReveal, Cyberlink’s PowerDirector 7 Ultra, Loilo’s Super Loiloscope, ArcSoft’s TotalMedia Theatre 3 Platinum, and Elemental’s Badaboom to evaluate how these applications utilize CUDA technology as well as what performance benefits consumers will notice during the video transcoding process. Each of these applications implements CUDA technology in different aspects of the transcoding process. Some use it for its raw computing power during transcoding, while others only use it for DVD upscaling and video effects. Each of the apps we chose will showcase CUDA in a different light and hopefully show this technology’s true versatility.
Next Page – System Configuration Testing Methodology