Why 99% of Power Supply Reviews Are Wrong

[Hardware Secrets Team]

There has been a new article posted.

Title: Why 99% of Power Supply Reviews Are Wrong
URL: http://www.hardwaresecrets.com/article/410

Here is a snippet:
"With computers (and users) asking for better power supplies, nothing more natural than reviewing websites publishing power supply reviews. But contrary to other hardware parts like CPUs, motherboards ..."

Comments on this article are welcome.

Best regards,
Hardware Secrets Team
http://www.hardwaresecrets.com

[jonnyGURU]

You know Gabriel, I was sweating bullets 'till the end waiting for you to find a way to beat me up.

Looks like I need to start using my variac for more than just changing input voltage from 115V to 230V (for testing EU only models) to impress you further.

FYI: I use a Sunmoon SM-8800. Have for about the last ten reviews or so. The SM-268 is in Matt Harris's posession and he uses it for Tech Gage and for reviewing single or dual 12V rail (since it only has two 12V load) for my site.

[Gabriel Torres]

LOL.

Thanks Jonny, I will update the article right now posting the correct load tester you have.

It was just that everything I said was stuck in my throat for some time... and I think in the throat of other good reviewers like yourself...

Gabriel.

[Gabriel Torres]

BTW, Jonny posted this interesting remark on his website that I'd like to share with our readers:

"Good to hear they're saving up for a Chroma. I like the Chroma, but they're too expensive, take up too much space and because they don't have actual ATX12V interfaces on them (just screw down terminals) there's some resistance added by the wires and connectors you have to run between your ATX12V interface board and the load tester itself. I know some PSU manufacturers use Chromas for the test reports they include with their PSU's and use them for some of the high load tests and for testing engineering samples because they can maintain high loads, stay relatively quiet and don't get terribly hot. But if you tour a PSU factory, you'll find the QC department has a bunch of tables with Sunmoon or TechRed ATX12V SMPS ATE's."

Enjoy,
Gabriel.

[jonnyGURU]

So you think you'll get a SunMoon or Techred now? Only about half the price of a similar spec Chroma.

In Chroma's defense....

The biggest problem with the "cheaper" units is that they don't have a built in O-scope (they have O-scope outputs already wired up specifically for ATX12V ripple/noise evaluation) like the better Chroma units. There's also no PC/software interface on the cheaper desktop units like the Sunmoons and the Techreds. Everything has to be recorded by hand.

But where they're typically used, in the QC room of the factory, they're usually testing for full load capability and voltage regulation and they're usually eye balling ripple on a CRT. No need for software, etc. The PSU either passes or it fails.

For reference:

Sunmoon is http://www.sunmoonate.com and http://www.sunmoontec.com

(No, they're not in English. When's the last time you saw a PSU factory in an English speaking country? )

and Techred is http://www.techred.com/_en_index.htm

[Andy]

So Im in IT now, but spent 10+ years as an electronic tech in the Australian Airforce and have civilian electronics qualifications.

To me, the idea of putting an external set of shunts, across which an oscilloscope(s) are then used to measure current provided to the test system is a flawed (in that its an approximation only) measurement solution. Flawed because the closed loop feedback mechanism used by the PSU is being applied to the shunt and the PSU under test in series. If the closed loop feedback was moved from the PSU output to the PC side of the shunt so that 12v in the 12v rail was applied to the PC instead of 12v minus the shut voltage then it would be more accurate. While the shunt voltage will be small, the reality is that when you are playing with big A numbers small V numbers still will calculate out to be significant power.

Of course moving the feedback point to the shunt wont be an insignificant task in itself, and Im not sure what the PSU controllers would do if the feedback point was accidentally taken to earth or left floating, both of which are potential stuff ups that could be made.

In my past, working with various ATE equipment it wouldnt be that hard to get some programable loads set up so that an appropraite GPIB controller / software solution could approximate a PC, assuming you knew what the load profile vs time looked like for your representative PC. All of which sounds a lot like the $50k monster your describing.

Andy

[Gabriel Torres]

Wow thanks Andy and thanks Jonny for your very important remarks.

Another friend of mine has also suggested me buying a load tester from this manufacturer: http://www.fastauto.com.tw

It seems that this is the load tester used by Seasonic and Corsair.

Well, it seems that now I have several other choices ;-)

I will take a look on all of them. I took a look on Sunmoon SM-8000 specs (although their website is in Chinese) and it seems a pretty good tool for me. What is its price range, Jonny?

Best regards,
Gabriel.

[ProphetCHRIS]

Fast is used by almost any manufacturer in China. They are quite ok but you still need to write down the data yourself.


The Pic I took in Topower's factory. Seasonic and Enermax are using the Chroma 8000 in their labs. There's just nothing easier and quicker for testing a PSU today. Jon just described it already that in the factory you just need to see if its "ok" or not.

[W1zzard]

very good input in this article. excellent read. the following is not meant to criticize your article but just to discuss a couple of things that have been bugging me re. psu reviews.

i am wondering .. how would you quantify a psu test result? at some point you have to say psu x is y better than psu z. or you dont?
what defines a good psu? i have a 350w psu here that can run fx60+2x x1900 xtx cf for $40 (its totally overloaded and hot but it works rock stable). is that good? is it bad?
then i see psus with 80% efficiency, yet their 12v line swings 0.5V between idle and loaded.

is the difference between good and bad psu "works" and "doesnt work"? any suggestions on how could we characterize the current loads that are required to pass "budget", "standard", "high-end" systems ?

anybody have experience how lower voltage stability affects the system if the system is stable at these voltage? any idea how overclocking capability is affected?

using a psu testing machine means that you test for a large number of changing variables. wouldn't this also mean that we should test video cards on an infinite number of system configurations? as far as i understand most reviews of most products give a snapshot of how the reviewer experienced the product in his unique test setup.

[reidy-]

Originally Posted by W1zzard

very good input in this article. excellent read. the following is not meant to criticize your article but just to discuss a couple of things that have been bugging me re. psu reviews.

i am wondering .. how would you quantify a psu test result? at some point you have to say psu x is y better than psu z. or you dont?
what defines a good psu? i have a 350w psu here that can run fx60+2x x1900 xtx cf for $40 (its totally overloaded and hot but it works rock stable). is that good? is it bad?
then i see psus with 80% efficiency, yet their 12v line swings 0.5V between idle and loaded.

is the difference between good and bad psu "works" and "doesnt work"? any suggestions on how could we characterize the current loads that are required to pass "budget", "standard", "high-end" systems ?

anybody have experience how lower voltage stability affects the system if the system is stable at these voltage? any idea how overclocking capability is affected?

using a psu testing machine means that you test for a large number of changing variables. wouldn't this also mean that we should test video cards on an infinite number of system configurations? as far as i understand most reviews of most products give a snapshot of how the reviewer experienced the product in his unique test setup.

I call BS no 350watt PSU can hold that kinda rig together you'r '40$' PSU isn't even going to be 300watt's which makes it even more like BS I want Pic's

you'r recommended for that kinda rig minimum of a real nice 500watt PSU

i bought a 650watt q-tec for my PC I only own an athlon 64 3500+ 2* 160gig HDD 2* DVD drives 1*7900Gs and a kn1 extreme mobo