NVIDIA CUDA-enabled Applications Roundup
Test system configuration and testing methodology
Test system configuration
To thoroughly test how these transcoding applications use CUDA technology in a consumer-level environment, we were careful to create a mid-range system that consumers could purchase at a relatively affordable cost.
Test system configuration
With this in mind, we put together a moderate AMD AM3-based system with 4GBs of RAM and an NVIDIA 9800GTX+ graphics card. Our eVGA GPU is factory overclocked with a 756MHz core clock and 2246MHz memory clock. It also has 128 stream processors which should make a huge difference in our benchmarks.
We would also like to note that we used NVIDIA GPU driver 185.85, which adds support for CUDA 2.2 for improved performance in GPU computing applications and expands GPU hardware acceleration for the NVIDIA video encoding library to GPUs with less than 32 cores. Here’s a complete run-down of the specifications of our AM3 test system:
- CPU: AMD Phenom II X3 720 Black Edition ($139)
- Motherboard: Gigabyte GA-MA790FXT-UD5P ($179.99)
- Video card: eVGA 9800GTX+ ($144.99 before rebate)
- RAM: OCZ Gold 4GB DDR3 1600 ($70.99 before rebate)
- Hard drive: Western Digital 160GB SATA ($59.99)
- Power supply: PC Power and Cooling 750W ($109.99 before rebate)
- Total cost before rebates and shipping: $700.95
The testing perimeters for evaluating the performance of these CUDA-enabled transcoding applications were three-fold:
- Evaluate CPU usage and determine how much of the computing load being handled by the CPU with CUDA enabled and disabled (Some of the apps do not allow CUDA to be disabled. In these instances, we will benchmark the app against our CPU-based transcoder to compare results.)
- Determine what advantages customers will notice using these applications with CUDA-enabled GPUs.
- What performance differences will consumers notice between using a strictly CPU-based transcoder and a CUDA-enabled transcoder?
After we determined our test perimeters, we also wanted a variety of video formats and sizes to choose from for our benchmarks. We choose everything from MPEG-4 and WMV to VOB and H.264 formats. This gives us a broad range of video formats that will likely appeal to a variety of consumers.
To evaluate the differences between CPU-based and CUDA-enabled transcoders, we selected a popular video transcoder called HandBrake. HandBrake is a multiplatform, multithreaded application that can handle a variety of different video formats and output all the resolutions we want to test in this roundup. HandBrake should serve us well in comparing the latest CUDA-enabled transcoders against a well-know CPU-based video transcoder. So, let’s get started.