I found a "white-paper" on TEC tech a week or two ago.
All About TEC (Go to the Peltier link in the left-hand column).
The gist of it seems to be this. Given the additional heat generated by the TEC cooler itself, they have graphed various combinations of heatsink thermal resistance (C/W) against thermal design power (load-level heat dissipation) of a processor. There are particular ranges of combinations -- heatsinks and processors -- where TEC is feasible and effective. But with TDP's in excess of 60W and heatsink thermal resistance below (perhaps) 0.20 C/W, it is totally useless to consider TEC.
Then, you add in the water-cooling alternative. This again makes TEC feasible with TDP's in excess of 70W or so (as one finds with the more recent AMD and Intel offerings). So you will find water-blocks manufactured with an integrated TEC hot-and-cold plate unit.
Again, though, there is another complication for this exotic cooling enhancement. The TEC device needs a lot of power supplied to it to move the heat ( with electrons) across the hot-plate-cold-plate boundary. You are not going to get an adequate supply of power from your PSU, or if you do, it is really going to put a strain on it. Further, some of these devices require a 24V potential difference as opposed to 12V. On manufacturer actually recommends buying two of a particular model PSU and stringing them up in series. So you are not only faced with the additional power consumption that these devices require, but a place to put the additiional PSU(s) that drive the devices.
Swiftech manufactures another device that includes a cooling loop for "chilled water" cooling of the CPU. The water passes across the cold-plate side of a TEC cooler. On the hot-plate side, another cooling loop runs water through the radiator -- removing both the TEC's heat and the heat from the water of the CPU loop. Again, there is the issue of powering the unit, and the simple fact that there are certain inefficiencies of introducing the TEC cooler in the mix. Even so, you can effectively reduce the CPU temperature below room ambient. I THINK that reductions to a level of a few degrees above freezing are possible with the TEC device applied directly to the water-block.
Again, because you are reducing CPU temperatures to near-freezing, or at least below room-ambient, there is the problem that the temperature will fall below the dew-point and cause condensation around the CPU socket, so you need to use a neoprene gasket to avoid the risk of water-droplets shorting out your precious components.
I think if i move to water-cooling, I will probably defer introducing TEC tech into the cooling loop. And I may defer it indefinitely -- longer than I seem to be deferring the move from air to water cooling.