AMD vs Nvidia Media Encoding
A lot of noise was made about using your GPU to do non-gaming related computations using the powerful shader hardware they contain. They are perfectly suited for highly parallelized tasks, and in many cases can outperform even the fastest desktop CPUs out there.
That said, there isn’t a whole lot you can do with your GPU aside from gaming these days. Some programs feature acceleration through the use of OpenCL. Most notable is Photoshop, but there are others as well. Both manufacturers are pushing for more extensive use of OpenCL – particularly AMD, since their APU products would benefit from it greatly when competing against Intel. Nvidia also has CUDA , but we won’t get into “CUDA vs OpenCL” here. They each likely have their own pros and cons, and discussing them is beyond the scope of this article.
Nonetheless, we will be looking at the non-gaming performance of these video cards in one of the first and most common “accelerated” tasks – video encoding. We take this test straight from our CPU test suite – ArcSoft Media Encoder 7.5, in which we convert a short 1080i clip recorded from a terrestrial broadcast. Results are recorded in the time it took to encode the video:
It looks like both video cards can encode faster than the i7 3770K, with the Radeon 6870 doing the best by quite a bit. It also beats QuickSync by the way; in this exact same test, the i7 3770K GPU finishes the task in 95 seconds.
One thing we need to consider is power consumption. We’ll look at power consumption in games in a moment, but for now we can take a quick look at how they do when they are being used for non-gaming tasks.
Interestingly, the GTX 560 requires a full power load when performing this task, something you might want to consider if you care about power effiiciency. It won’t do you much good to save 25% of your encoding time when you are spending 250% more power doing so.
For OpenCL testing, we’ll be using SiSoft Sandra 2012. It has a built-in test that uses OpenCL in pure software mode on both GPUs and CPUs. This might give us an idea on how programs like Photoshop benefit from these video cards:
Throwing out the CPU score (since it is not intended to run this code), we have a virtual tie here. Interesting, considering the gaming performance is also pretty close.