The K2000 with a 128 bit wide bus, 384 cores and only 64 GB/s memory bandwidth is left in the dust by the GTX 680 with 256 bit bus, 1536 cores and 192 GB/s memory bandwidth. Yeah, the GTX consumes more power, but it is way more powerful. You can't compare the performance and fuel consumption of a Fiat 500 (K2000) to a Mercedes S500 AMG (GTX 680).
Keep in mind that Quadros are nVidia's cash cows, overpriced and underperforming, but they get away with that because of superb Marketing (= spin-doctors who can convince nearly everybody that the shortest distance between two points is not a straight line, similar to CC Marketing).
Quadro K2000 vs Geforce GTX 680 Ray-Traced Renderer
Geforce GTX 680: 6 minutes and 11 seconds
Quadro K2000: 28 minutes and 11 seconds
@Harm: Ok thx... that's what I was thinking
@Terry: The benchmarks are with the old Quadro... not the new Kepler technology... the K2000 has double the ram of the old one... so the render should be at least double the speed, I guess?
The main performance determining factor for a video card is the memory bandwidth, not the amount of memory. Memory bandwidth is more important than number of cores too.
See a limited performance comparison here: Video Card Performance and scroll down a bit till you see the relevant graph.
Interpolating, one would expect the performance of the K2000 to be between the GTX 550 Ti and the GT 640, possibly with a score around 120-130 seconds, around 4 times slower than a GTX 680.