Today’s high performance motherboards are able to take advantage of the graphics processing unit (GPU) built-in to the graphics card in order to assist the CPU in completing processor intensive tasks. Known as parallel processing, or stream processing, this method of employing the GPU for applications other than graphics processing is now a standard for all top motherboards. Each of the major CPU processors have their own proprietary architecture for distributing processor load between the CPU and GPU. Nvidia’s parallel computing architecture is called Compute Unified Device Architecture (CUDA), while Intel processors use Quick Sync technology and AMD processors use ATI Stream technology. The premise of this technology is the same, but the performance differs for a number of reasons. This review focuses only on the performance of these parallel processing architectures.

Test Equipment

These numbers are based off a test run by TomsHardware.com contributors in January 2011. The reviewers compared an Intel HD Graphics 3000 system with an Nvidia GeForce GTX 570 system and an AMD Radeon HD 6870 system running CyberLink MediaEspresso to convert a four minute 1080i video to 1024x768 H.264. A test was also perfomed using Arcsoft MediaConverter 7While there may be other extenuating factors that affect the performance times, the processors selected here are comparable.

No Acceleration

Before you can understand the benefits of parallel processing, it’s best to take a look at how these computers perform with no acceleration. According to the tests, it took a machine without hardware acceleration 2 minutes and 52 seconds to complete the video conversion. With ArcSoft MediaConverter 7, the no hardware acceleration machine clocked in at 1 minute 35 seconds and reported 30% CPUO utilization.

ATI Stream

The AMD Radeon HD 6870, which uses the ATI Stream technology, recall, took 1 minute 26 seconds to complete the process. With Arcsoft, the AMD Radeon HD 6870 completed the process in 1 minute 8 seconds. Interestingly, Arcsoft reported that the AMD Radeon HD 6870 had a lower CPU utilization than the other processors.

Nvidia CUDA

The Nvidia GeForce GTX 570 completed the Cyberlink test in 1 minute 23 seconds and the Arcsoft test in 1 minute 16 seconds. It utilized more CPU than the ATI Stream process and the no hardware acceleration scenario, measuring up at 50% CPU utilization.

Intel Quick Sync

The Intel HD Graphics 3000 machine completed the Cyberlink performance test in an amazing 22 seconds and the Arcsoft test in 41 seconds. By far, the Intel HD Graphics 3000 machine is the fastest when compared to the Nvidia CUDA process and the ATI Stream technology used by the AMD Radeon HD 6870. Reports on CPU usage weren’t available.

Conclusion

The CPU utilization figures are somewhat telling. Bar none, the Intel processor with Quick Sync is faster. But the AMD Radeon HD 6870 has the lowest CPU utilization, meaning that more of the work is likely being done by the GPU. The Nvidia GeForce GTX 570 has good performance, but higher CPU utilization. Bottom-line: for speed, go with the Intel. But the ATI Stream technology may be able to kick more of the work to the GPU, meaning that if you have a high end graphics card, you may be able to benefit from an AMD motherboard.