Adobe Mercury Playback NVIDIA only Legality

[Walter Soyka]

The new Kepler cards are built on a different architecture than the previous generation of Fermi cards. CUDA core counts are not directly comparable between the two.

Walter Soyka
Principal & Designer at Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog – What I’m thinking when my workstation’s thinking
Creative Cow Forum Host: Live & Stage Events

[Drew Hoover]

The CUDA core counts are not comparable because the Kepler architecture (I’m getting a little out of my element here) eliminated the second (shader) clock in the Fermi cards. A unified clock speed means that everything is more efficient, but more cores are required. So blah blah blah, more CUDA cores does not mean better when comparing GTX5XX to GTX6XX.

You are overlooking some very important features that add a lot of value (or don’t, depending on your needs) to the Quadro.

1) Quadros use Error Correction Control (ECC) memory. Basically ECC memory is extremely reliable (and expensive) and doesn’t allow for errors when you’re say, applying a color grading effect to a clip. The card isn’t designed to be a speed demon, it’s designed to be accurate. No one cares about a compression artifact in one frame while playing Battlefield 3, so GTX cards don’t have ECC.

2) Quadros are supported with really specific drivers that enable them to work well with all kinds of 3D design programs.

3) Workstation GPUs are single slot cards, so you can easily stack a bunch into one server.