Peaople have been talking about that for some time, but no serious Photoshop user is actually doing this.
Have you made any tests that conforts you in the fact that Photoshop is actually faster (there are some tests using actions around on the internet). I did some, using the RAM disk or not and didn't notice much difference, and there is a good reason for that.
If you actually watch your RAM use during Photoshop work (CS3 in my case, my OS had always been up to date after testing on a test machine that my work softwares had no issues), you notice that all of the RAM (whether it is 8GB or 16GB or whatever) gets filled as inactive RAM when you work on large images (I work on images from very small up to tens of GB, biggest ever opened and worked on 94GB). This is RAM being actually used but can get free up when another application is open or in need of some. OSX uses this as swap, before writing on disk. So Photoshop first fills its allocated ram, then sends disk io requirements to the OS, the OS fills its RAM with that and then Scratch disks are physically used. This is the simpler way I found to describe this. As it is simple it is not exactly the truth but not far.
I could conclude that making ram disks doesn't allow Photoshop to be faster than letting OSX manages the RAM the way it does.
Also please note that Scratch disks are ALWAYS used. Whatever is done is always mirrored on scratch disk. Available ram just helps by giving head room.
If any doubt go search for Ramon Castaneda's posts on Adobe Forums. He knows all this better than all of us.