Too many power supply manufacturers (hundreds if not thousands) meaning too much time to assemble a list. However basic market techniques used to buy anything technical also apply to a power supply.

First, market retail price makes obvious supplies that are missing essential functions. Minimum sufficient supplies are about $60 full retail (not to be confused with discounted prices for that same model). That does not even imply a $100 supply is sufficient as one might assume using binary logic.

Second, a manufacturer provides a long list of numeric spec. A manufacturer dumping an inferior product will not provide those specs so that the 1% who actually know technology cannot 'blow the whistle'. What constitutes numeric specs for a power supply? So many numbers as to fill a page. Some of those manufacturer numbers that must exist in writing would include:
Short circuit protection on all outputs (short all outputs together and still power supply will never fail)
Over voltage protection (power supply will never damage any other computer components)
Over power protection (too many computer components never causes damage to a supply)
EMI/RFI compliance for CE, CISPR22, FCC part 15 (ie put an AM (medium wave) radio next to computer and have no interference)
Safety compliance for VDE, TUV, D, N, S, Fi, UL, C, UL, CB
Full load hold up time (how long power supply works just fine with no AC power input): 16msec.
Efficiency greater than 65% (power supplies should exceed this number although conservatively rated supplies may only list this number)
Ripple or noise 1%
MTBF under full load at 25°C >100k hrs
Acoustics noise listed in dB while under full load.

Third, only useful power supply reviews always do tests such as loading the power supply to 100% rated power and measure its voltages. Also short all power supply outputs and demonstrate no failure. Tom's Hardware was one of so few web sites that performed such tests. Most reviews are so technically pathetic as to discuss appearance, weight, and some implied convenience.

Fourth, recommendations from many computer techs often mean nothing. To be A+ Certified, a computer tech need not even know how electricity works. Too many techs will push irrelevant numbers such as total wattage and will also hype more fans.

Even better supplies are also universal. These are features beyond minimal requirements. For example, a 120 volt power supply works just fine when AC voltage is at 90 volts or rises to 265 volts - just like all laptops. Many supplies that do this will claim a minimum voltage of 100 VAC or less. Such 120 VAC supplies will often work even at 85 VAC.

Even better supplies also include power factor correction which means its AC power interface is even more robust as well as more efficient.

Some manufacturers are often listed repeatedly for their superior designs including Antec, Computer Power & Cooling, and Seasonic. That does not mean every supply from these companies are best. But as noted previously, no supply observed in CompUSA appeared to be minimally acceptable. Dumped supplies into an American market flooded with technically naive computer assemblers is so profitable.

A quickest identifier of a defective supply is the AM (medium wave) radio test. If a power supply interferes with reception, then the power supply probably has multiple severe deficiencies.

Move on to wattage. Total wattage says little useful. For example a Dell 300 watt power supply might be rated as 430 watts by a clone manufacturer. Neither watt number is a lie. But many computer assemblers would not understand why those two numbers accurately measure the same supply.

More important are output amperage for each voltage. Unfortunately, no one can accurately say what the actual load will be or if that power supply is actually performing properly. A fully assembled computer with all peripherals accessed simultaneously has output voltages measure using a multimeter. If building a computer, there is no way around confirming hardware with measured numbers. If the computer has a motherboard voltage monitor, that motherboard function also must be calibrated with the meter.

None of this is complex. Most of this takes longer to read here than to actually perform - except if a power supply manufacturer makes doing this difficult.

Wow, thanks for that informative answer.

One more question, which may sound kind of dumb after all the information you just provided, but the fan in my 2 year old Dell at home makes a fluctuating noise. When I'm doing something really CPU intensive it makes a lot of noise and when I'm just typing in a reply box in the Cellar, it makes a sort of oscillating noise. Sometimes louder, sometimes softer. The PC seems perfectly fine otherwise. This noise isn't a screech like bearings failing, it's maybe just the fan speeding up or slowing down. Is that in itself a signal that the power supply may be failing? If it's failing, I'm going to be glad that you posted what you just did, so I can get a decent replacement.

P.S. I have a multimeter that I use for household electrical work, but have never used on electronics.

That's probably not a problem. In the past few years some manufacturers leave the fans low when they're not as necessary, so the PCs stay quieter.

As UT notes, a power supply may change fan speeds based upon temperature. These design variations are numerous. Early versions has a thermistor inside the fan alongside a transistor and hall effect sensor. Yes, electronics are even inside each fan.

A computer assembler really cannot duplicate this. That power supply fan must move enough air to also keep the case cool in a 100 degree room. IOW that fan control must be integrated with other electronics on motherboard. Dell can integrate that fan control. Same may be too complex for a computer assembler for a clone machine.

Easier is to integrate under motherboard control a fan that often makes creates noise - CPU heatsink fan.

CPU temperatures can vary drastically. To appreciate how much, view why that power supply is located on motherboard adjacent to the CPU. A CPU may go from drawing less than one amp to tens of amps. To make a CPU even hotter, that current demand may occur in microseconds. Therefore a CPU fan may power cycle because CPU power consumption can vary greatly.

Another sidebar - why do fans fail? Many will blame only what they can detect - a sticky or worn bearing, or excessive dust. Fans are designed with sufficient torque to overwhelm such restrictions. However, if fan electronics sensor gets misaligned, then actual fan torque can be signficantly reduced. Some fans may not startup everytime or may eventually start moving later. One such fan over a generation ago resulted in three separate service calls involving a long airplane trip before a mysterious failure was finally traced to a 'sometimes sticky' fan.

I'm not going to worry about it then, since there are no other symptoms. Thanks, both of you.

Ok, I have been doing some playing, I haven't raised the funds to get a tester just yet. However on closer inspection I have notices that my cpu fan is not spinning and my hard drives are spinning.

Don't run it in that condition. The CPU fan is necessary.

It can be a power supply, but I've had dust cake onto a CPU fan and stop it when it got a little damp. The dust turned to paste.

It can also be that the CPU fan has become unplugged from the mainboard fan header. Look for the two wires coming off the fan, and make sure they're plugged in somewhere.

The tester is useless. Get a 3.5 digit multimeter - a completely different item - as sold in Sears, Radio Shack, Tru-Value Hardware, K-mart, Lowes ... A tool that will report on all components is a power supply 'system' - more than just a power supply. A tool that will provide numbers so that others can add additional assistance. Tester does none of that.

Best price (under $20) for this commodity tool will probably be in Wal-mart. Required tool is sold as a commodity - much like screwdrivers. Tester is a unique (exotic) tool that does nothing useful for your problem.

If CPU is Intel, then a seized CPU fan would cause no harm.

is the mulimeter similar to an Ohm meter.

Random note: the CPU is AMD, not intel here

Multimeter measures ohms, volts, and amps. It displays a measurement with at least three digits. You would be using its DC Volt function; 20 volt range.

If CPU fan is not working, still, CPU would at least boot. If AMD processor does not overheat too fast, then damage would not occur. It would just crash.

However this is getting too far ahead. First get voltage numbers.

Attach black lead to chassis or push the probe into the nylon connector to contact any black wire (from power supply). Touch the red probe to purple wire. Read and record that voltage both before and when power switch is pressed. Do same for green and gray wire. Later, do the same thing for any one red, orange, and yellow wire. But first, learn what those first three wires report on the meter.

I must make money to get meter now

UPDATE:
I got the power supply replaced...
Still down checking motherboard and new power supply

Well you could not find $20 to by a multimeter, but found $60 to buy a power supply? Then you did exactly what was recommended as not doing? How do you expect to get the computer fixed when you follow your feelings rather than facts?

Meanwhile, even a defective power supply can boot a computer. How do you expect to confirm power supply integrity again without the meter?

Replacing a power supply based only upon wild speculation is, well, ... you tell me. Exact same reasoning proved Saddam had WMDs or that it was safe to launch seven Challenger astronaut to their murder.

You still need a multimeter if fixing the computer is your intent. Appreciate the value of what you just did. The value in making mistakes - we learn from them.

No I finally got a hold of a friend who does this sort of thing and he ran the tests and replace it.

Oh. So its fixed. I did not understand that part.