I quite like using TINA-TI. I wasn't aware that they seem to be trying to transition to the Cadence/Orcad simulator - I'll have to download it and see what it's like.
I think I'm right in saying that the backend for these is the MIT SPICE libraries, so in general terms they'll all produce similar results (and have the same limitations) - it's mostly a case of how you get on with the GUI (for some reason they're all a bit quirky and odd, though you quickly get used to enough to do the basics). However, manufacturers do extend them a bit, so if you're working with a particular vendor's products, there may be a case for using their simulator to get the best out of their models (that's certainly the case with LTSpice).
Unfortunately, most of these are still stuck in a Windows-only world, though apparently LTSpice runs under WINE.
Helpful hint: make sure you have a ground on your schematic and that there's a DC path to it from any node on your circuit, otherwise the simulator will throw a wobbly when it does the initial DC analysis to get the operating points for the AC part of the simulation. (A node being any wire that you draw to connect pins.)
If you want to understand why, consider two capacitors in series. That's quite valid as a physical circuit, and will behave as though it were a single capacitor. But what is the voltage in the isolated section between the two capacitors? It could be anything, depending how much free charge was trapped there when the capacitors were joined, so the simulator has no way to put a figure on it. In a real circuit, even the very best capacitors have a small leakage (and the dielectric would break down for very high voltages), so the node eventually ends up close to circuit ground potential from the periods when the equipment isn't powered, and you'll never see millions of volts there. That leakage is the clue how to deal with it in the simulation: include a high resistive path around one of the capacitors. In TINI-TI, you can right-click on the capacitor part and tick a box to give a 1G path in parallel with the capacitor. That's a nice way to deal with the problem, because it's hidden from view and you don't end up with a schematic littered with extraneous resistors, but if you were doing something really critical and precise, you might instead want to use your own higher-value resistor.
