Bit or Memory??

The clock rate of a graphics card is the overall speed at which it can retrieve and calculate the data to push to your monitor. At low resolutions the higher clock rate can be noticed with extremely high performance. These differences are indicated in low resolution to see the performance gains, cause like a CPU it's impacted mostly on raw-data calculations.

The memory essentially allows more data to be sub-sequentially stored on the card for calculations. Which helps eliminate latency and bottlenecks at higher-resolutions with significantly more data being transmitted. These differences are indicated in high resolution environments, cause like your machines memory the more applications open cause latency as the memory HEAP becomes full...

The bits are simply referring to how much data at any one time are being transmitted and calculated. The more bits the more congruent colors and effects. An example would be if you have 32 bits, that's eight bits each for your RGB and Alpha scale.

Higher in all three is obviously better, but the primary focus should be clock and memory. Another thing to keep in mind is that lower dye size and architecture sizes also can impact performance. An example would be Nvidia's Fermi and Kepler architecture; Kepler is focused on multiple streams of data at any given time. Where Fermi is single task oriented; that is why a Kepler provides such a significant boost over it's predecessor.

That is why higher end cards usually have higher values in all three categories.

Hopefully that answers your question, if any part of it is incorrect please let me know. It's been awhile since I've been in my Electrical Engineering class where we went over this.