OpenCL: Video Processing (Basemark CL)
We see a lot more green bars toward the top of our video processing charts. The video filters used here are quite similar to those used on the previous page's image processing tests. This time around, however, AMD's cards don't perform as well (with one exception) due to their weaker video hardware. The only discipline in which AMD pulls far ahead is the compute-intensive surface smoothing subtest. All three implementations of the GeForce GTX 660 Ti offer exactly the level of performance we have come to expect from them over the last few pages. They also prove to be strong alternatives to the GeForce GTX 670.
Everything we've said about the Kepler architecture in past stories applies to GeForce GTX 660 Ti as well. It’s a great architecture for gaming, but tends to underwhelm in compute-intensive workloads with very few exceptions. This is expected, since Nvidia’s employs the same GPU as GeForce GTX 670 and 680.
If you’re focused on gaming and your favorite titles don't employ a lot of DirectCompute or OpenCL elements, then the GeForce GTX 660 Ti is a reasonable choice. Looking to the future, however, Nvidia has to hope that Metro 2033 isn't an indication of what developers are planning to do. That title's depth of field filter requires enough compute power to bring even a GeForce GTX 680 to its knees.