Actually, I found one more power source scheme and that cannot have -12V rail strengthened easily as I proposed. So the idea is not that universal. I do not know which kind is more common. My
guess would that the one which can have -12V strengthened. Anyway, it looks like one needs to check how his/hers particular PC power source is connected at the secondary side.
In general, off-the-shelf 24V power sources are easily available as 24V switching power sources for LED lights. They seem to top at about 350W which should be enough. I only consider adjusting PC power source since me (and probably almost everybody else) has some spare one from older PCs.
I use two small (SFX format) PC power supplies in series to get 24V. One of them is floating ground of course.
Most hotends use 3W wire wound resistor of 6.8Ω, axial leads, 13mm length, 5.6mm diameter. We mostly power it with 12V which gives us power of 12²/6.8 ≅ 21W. Much more than the nominal 3W, but it does not really matter. The point is that resistor power rating is limited above all by how quickly it can get rid of heat. That is the reason the higher power resistors are bigger in size. Since our hotend resistors are surrounded by thermally conductive paste and inserted in a metal block they can get rid of heat much better than when they are simply soldered to a PCB. On a PCB it would have maximum 3W, in our hotend it can withstand 21W easily.
I actually power this 6.8Ω resistor with 24V which leads to maximum power of 24²/6.8 ≅ 85W. And I did not have any problem with it yet. But notice this is the peak power. I use PID and limit the duty cycle to at most 120/255. That means that my maximum average power is about 86/255*120 = 40W. Which is just the power the guy mentioned. By the trick, he probably meant two things preventing baking our 3W resistors to ashes):
1) good heat transfer from resistor to hotend
2) limiting PID duty cycle if more than 12V power source is used.
Some people go from hotend resistor to heaters. This way they make sure that they really do not go over nominal values. I do not think it is needed. Wire wound resistor should be enough and they are cheap. Just make sure that heat is transferred well from the resistor to the hotend. Although I'm thinking about replacing the hotend resistor with a 15Ω one to limit peak currents and a chance they will induce something to thermistor wires. But I'm not in a hurry. The high currents I have there now do not seem to do any harm. So the replacement is just a quest for unnecessary perfection
I'm not considering going for a heater.
The 40W cartridge heaters, if specified for 12V, should pull 40/12 ≅ 3.3A. They should have resistance about 3.8Ω. I do not think is it that great idea to use them, since they are more expensive. They will heat up your hotend sooner. I already achieved that by moving to 24 V, which gives me also higher speed on steppers (but I care only about extruder stepper speed).
Other advantage, I can think of, is that they are run on nominal values so they stay in warranty. I never heard of hotend resistor failure yet and it did not happen to me, so I think cartridge heaters are just vanity ... at least till I do not find some data confirming their usefulness
Any wire wound resistor which just fits nicely in your hotend should be fine. If you care much find surge wire wound resistor (these can withstand short current peaks better) and look for the ones which are rated for the biggest temperature.
Edit: spelling.