Well it wont hurt to try. The way I see it is if NVidia got the silicon manufacturing process down to the point where defective units were not common enough to keep up with the demand for 670's, they may have used units that pass the qualifications for a 680 on the 670 boards. I work for a equipment design group and most of the assembly lines we make have a unit production output of about 1 unit every 8 1/2 seconds, with a failure rate of about 1 unit for every ~6000 units (and every unit is tested for quality control). Some of the lines we have made are designed to change product features on the fly, usually producing enough of one product to fill our retail demand for the next quarter, then changing the settings for a version with more/less features. I can't say what this product is, but I assure you many of you have one, and frankly I'd be screwed if I told you that you can upgrade to the better device with a firmware flash and a couple pins on the CPU tied low.
But point is, I want to see if all nvidia did was just disable stuff with this method for the GPU's that are 680 spec, but intended for 670 boards.