Why 99% of Power Supply Reviews Are Wrong

[Keh]

Having stumbled upon this tread, I was impressed enough to register so I could toss in my 1.65 cents worth:

1. The use of a multimeter is a declaration that the person has no idea of what they are dealing with. Multimeters sample at their designed rate and display an average which completely masks the underlying reality of wildly shifting loads. (If this were not the case, multimeters would have a blur for their least significant digits.)

2. Voltage is EVERYTHING. The current don't flow 'til the voltage says go. Current is merely the result of a voltage differential and is limited by the resistance to it.

3. The 'resistance' (readers please note the quotes before madly posting that there is more than simple dc resistance in play) presented to each rail of the PSU is a maelstrom of shifting values. One moment just a tad, at another, a gargantuan jump. It is to this nightmare that the PSU must react. Testing with static loads is a joke.

4. The specs for PSUs are NOT expressed as 'averages' - if your voltages show a sag, you're 'near the edge', if not 'in deep doodoo'! If the voltage(s) spike/sag outside the limits, ALL BETS ARE OFF.

5. PSUs are mostly 'snake oil' - sold to the indifferent and the gullible (not to mention the adamantly ignorant). I remain convinced that a 'cheap' PSU costs $2.95 to manufacture while a 'quality' unit costs $4.95; the rest is packaging, transportation, advertising and mark-up, mark-up and more mark-up.

6. The 'rating' of a PSU is pretty well WHATEVER the vendor wants it to be; they're the ones who set the parameters for the derivation. (I remember some 30-odd years ago reading an article showing how a stereo amplifier that had true specs of 5W/channel, 20-20K +/- 0.5%, THD/IMD under 1.0% could be advertised as a 200W/channel unit - LEGALLY!) The soul bragging about his 650W Q-Tec has yet to discover that it is 'rated' in peak (transient) power, not continuous (where it would be lucky to hit 425W) and that that particular brand has garnered such a bad rep that it has been abandoned.

7. PSUs are made with 'industrial' grade components (+/- 20%). The actual capacity of an assembled unit will fall in a classic bell-shaped pattern. The bean-counters are the ones who decide the 'most profitable point' between 'returns' and 'reputation'. (Anyone still believe ANTECs are 'built like tanks'? There was a time...)

8. The 'average' system draws less than 200W from the wall socket. Those who cannot grasp the difference between 'average' at the wall and instantaneous 'peak' (all rails SIMULTANEOUSLY) at the PSU output will continue to live on their 'luck of the draw', convinced that a 450W unit can run anything.

I doubt I will ever see a PSU evaluation that actually 'tests' the PSU's ability to maintain voltages within spec (down to the microsecond, or at least millisecond) while each rail was being jumped (both up and down) by 1,2,3...max amps from each possible point (1, 2, 3...max amps), SIMULTANEOUSLY at RANDOM rates. But then, I doubt there is a PSU that could pass.

And considering that component manufacturers are no more forthcoming in their actual peak demands, it probably doesn't matter.

Caveat emptor.

[reidy-]

^^ Great post and a pretty interesting read, and don't you knock my q-tec lmao yeh I know its duff (@650watt's they are renown'd to hit 50% effeicency ) but it was a b-day present and I am way to skint too afford an hiper 580.

as long as it holds together when running my athlon few HDD's x1950pro and a new mobo (old one let go just a week ago ) I shan't complain...much


,Reidy-

[Gabriel Torres]

Hi,

Thanks for all the input posted.

Regarding the question on how to say power supply A is better than power supply B, my opinion on that is the following.

It all depends on the cost/benefit ratio. I.e. the price against the power the unit can deliver, plus extra features.

For example, you cannot say a 700 W brand A unit is better than a 400 W brand B unit simply because they are targeted to different users in mind.

But when comparing several power supplies at the same power (or price) range, things get easier, and then you can say which unit is the best based on several aspects the reviewer should declare upfront, like cost/benefit ratio, number of extra features, temperature, how much real power the unit can deliver, etc.

Each publication has its own methodology, so if you are planning to create your own methodology you should write down what is most important for you and publish what is the method you are using to classify the results.

Cheers,
Gabriel.

[jonnyGURU]

Originally Posted by Keh

5. PSUs are mostly 'snake oil' - sold to the indifferent and the gullible (not to mention the adamantly ignorant). I remain convinced that a 'cheap' PSU costs $2.95 to manufacture while a 'quality' unit costs $4.95; the rest is packaging, transportation, advertising and mark-up, mark-up and more mark-up.

6. The 'rating' of a PSU is pretty well WHATEVER the vendor wants it to be; they're the ones who set the parameters for the derivation. (I remember some 30-odd years ago reading an article showing how a stereo amplifier that had true specs of 5W/channel, 20-20K +/- 0.5%, THD/IMD under 1.0% could be advertised as a 200W/channel unit - LEGALLY!) The soul bragging about his 650W Q-Tec has yet to discover that it is 'rated' in peak (transient) power, not continuous (where it would be lucky to hit 425W) and that that particular brand has garnered such a bad rep that it has been abandoned.

7. PSUs are made with 'industrial' grade components (+/- 20%). The actual capacity of an assembled unit will fall in a classic bell-shaped pattern. The bean-counters are the ones who decide the 'most profitable point' between 'returns' and 'reputation'. (Anyone still believe ANTECs are 'built like tanks'? There was a time...)

No offense, but portions of your post makes it sounds like you're bitter and ignorant (I mean that in the nicest way possible.) Perhaps with an axe to grind?

The caps in a PSU alone will cost the OEM at least $5 and double that if they use Chemicon, Hitachi, etc. $3? $5? That's laughable. Just the wiring without connectors and sleeving costs about $2 per unit at the factory.

As for the specs, you need to read Gabe's power supply "reviews" more often. A PSU is comprised of components that can dictate how much power a PSU can put out. Not saying that some PSU manufacturers put whatever label they want on some PSU's, but that's the exception not the rule. A good example of this is the Youngyear that Gabriel just had a look at. Judging by the components, the unit likely would not put out 350W. But they label it 550W.

As for Antec, at least they use components rated for the outputs they claim. Unfortunately, the CWT built units fall vicitim to volatile capacitors (Fuhjyyu) and poor cooling. A near deadly combination that may render a PSU useless in a years time when used in heavy battle. But that's not to say that they're cheaply built or are incapable of doing what Antec claims they can do.

And Gabe, my SM-8800 costs about $5K.

[Gabriel Torres]

Wow, thanks a lot for your post Jonny. I was afraid to reply to that user without sounding harsh.

Regarding your SunMoon unit, a lot of people recommended that one as well, thanks for posting the price.

Also thanks for supporting us on your website. I simply loved the post you made about that Young Year...

Cheers,
Gabriel.

[Keh]

Hmm, guess using hyperbole to make a point was a bit of a risk...

Still, if caps alone are $5, one is hard-pressed to explain all those $20 "500W" PSUs we all see being flogged hither and yon. (Reverse hyperbole?)

Bitter? Like to think more along the lines of 'experience' and 'observation'.

There are really very few actual 'manufacturers' of PSUs - most offerings are 'rebrands' where the vendor's beancounters have either selected a manufacturer who specializes in servicing 'brands' that have no concern about staying around for long (they can always invent a new 'brand'), or one that is willing to compromise on component quality/sizing to get the contract.

The 'rating' of a PSU is solely dependent on the 'assumptions' behind the rating. (That is how the 5w/ch amp ended up with a claim of being a 200/ch unit.) You can set those assumptions very conservatively and have a PSU with good 'headroom' or you can set them very loosely and have more complaints and a lesser reputation (but much greater profit). Five years ago, I would not hesitate suggesting ANTEC (and they were CWTs); today I would not think of it (and today, not all ANTECs are CWTs).

'Snake-oil' salesmen thrive in a market of ignorance. I have never been able to reconcile the fact that most computer 'enthusiasts' will delve into the most esoteric aspects of video and audio cards, chipsets, etc, but assume that all that needs be known about a PSU is the 'number' in the title.

Worse are those 'power calculators' that use values derived from heaven knows where and when. (And even if actually asked, is a component manufacturer going to give out the 'peak' demand or the minimal 'average' or perhaps the minimum?)

But my favourite are the souls who believe that they can plug their system into a wattmeter and use the reported value to determine their PSU needs.

And no-one ever thinks about the impact of the quality/capacity/efficiency of the mobo's regulators in the equation...

It is in this environment that PSU vendors compete - a recipe for a 'race to the bottom' (or gouging those who believe that the 'bigger the number', the better.)

This article was one of the few on the subject that deserved support.

[Gabriel Torres]

This article was one of the few on the subject that deserved support.

Really thanks for your comments man. It really added something to the discussion.

Cheers,
Gabriel.

[rarchimedes]

Gabriel,

This is a minor niggle, but precision is of interest. You note that switchers work in khz, which is basically milliseconds, while responding in microseconds, which is millionths of a second. Whichever you choose to represent it in, the response is in the same basic order of magnitude as the switching speed. Using two different terms tends to confuse the issue.

I was in the industry when switching power supplies first came on line(early '80's), and they were an absolute disaster in their early incarnations, because they often failed as a short to ground. In our labs, I dived at a couple of machines to shut them off as the grills on the power supply started to glow. These were large systems drawing significant amounts of power, so it wasn't easy to just fuse them, and emergency shutoffs had their own dangers in large, largely 3-phase systems. The switchers disappeared for a while, and then had a resurgence as personal computer sales took off and the designs improved. The real key to any modern power supply is the capability and quality of the capacitors. In the early '80's, many of the modern capacitor types just did not exist.

[Gabriel Torres]

Thank you so much for bringing this to my attention. I will revise the text as soon as the time allows.

Cheers,
Gabriel Torres