Article on PSU size & overkill; irresponsible journalism

mcoleg · Post by **mcoleg** » Sun Jun 10, 2007 10:05 pm

http://www.overclockers.com/articles1452/

quoting power supply expert jonnyguru: it's "irresponsible journalism"

http://www.jonnyguru.com/forums/showthread.php?t=2443

Das_Saunamies · Post by **Das_Saunamies** » Sun Jun 10, 2007 10:24 pm

I'd call that responsible above all. Case in point for green and reasonable computing.

Only the PSU manufacturers, if even them, know what the PSU can and can't take in the long run. Power efficiency does fall over time, but the pace is dictated by the quality of components, which could be improved instead of resorting to brute force(kW power) methods. And over 50% consumption? Doubt that occurs even 25% of the time if you enable the proper features.

I find the article an interesting experiment that proves, to an extent, a point in favour of reasonable wattage and component quality: there's no reason to just throw more of everything at the consumer, one could simply improve what is already there to match the standards expected in modern PCs. I don't know for certain how much more of everything goes into a high-power PSU, but the weight differences suggest more material gets used... and wasted! I think most of us could lead a happy PC life with a high-quality 400-500W PSU that lasted us for 4-5 years, efficiency loss included, saving many a components from being produced and dumped in vain. The manufacturers would have an easier time as well, since there would be no (insane) race to new heights in W readings to win market share and bragging rights.

I didn't quite find mention of your method for measuring the power consumption, voltages and amperage though. There was mention of an S&M PSU Test but I have no idea what that is or how it works.

mcoleg · Post by **mcoleg** » Sun Jun 10, 2007 10:44 pm

power consumption and amperages measured with kill-a-watt. not the best method since there's no data logging but for peak numbers it's ok.

voltages are measured with a radio shack's voltmeter that does have data logging. they cut the 12v rail graph from the article.

more extensive info is here:

http://www.ocforums.com/showthread.php?t=511180

you have to browse in, sry. there was more experiments though - for example i applied some heat to the intake of the psu to pry to decrease the efficiency and so on.

here's the original draft:

http://mcoleg.blogspot.com/2007/05/all- ... -that.html

Das_Saunamies · Post by **Das_Saunamies** » Sun Jun 10, 2007 10:56 pm

Okay, thanks for the clarification. "Kill-a-watt" and "Radio Shack" are very foreign concepts for me, I've only come to know the latter through internet forums and the first sounds like something out of Warhammer 40k.

mcoleg · Post by **mcoleg** » Sun Jun 10, 2007 11:03 pm

oups, sry

i didn't find the multimeter on their site but it's close to this model:

Multimeter

kill-a-watt:

Killawatt

you should have similar gadgets in europe; "Energie Monitor" or "Energy Check" by Voltcraft; may be more that i don't know about...

Admin edit for side-scrolling URL.

jaganath · Post by **jaganath** » Mon Jun 11, 2007 12:06 am

quoting power supply expert jonnyguru: it's "irresponsible journalism"

what total BS. if anything is irresponsible it is the misleading of whole legions of gamers into thinking they need a PSU 100-200% larger than they actually do.

it's only going to tolerate being run at 50% or higher loads for so long.

also BS. a quality PSU with a well-engineered cooling solution will tolerate loads within its rated capability for at least as long as the average PC lifetime (5 years?).

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 4:24 am

jaganath wrote: PSU with a well-engineered cooling solution will tolerate loads within its rated capability for at least as long as the average PC lifetime (5 years?).

do you have some data on that? some links? i've been trying to find something but so far unsuccessfully...

Techno Pride · Post by **Techno Pride** » Mon Jun 11, 2007 5:23 am

jonnyguru: give it one year, it won't last that long.

bush: give me one year, osama can't stay hidden for long...

hehe couldn't resist.

Mr Evil · Post by **Mr Evil** » Mon Jun 11, 2007 5:27 am

mcoleg wrote:
jaganath wrote: PSU with a well-engineered cooling solution will tolerate loads within its rated capability for at least as long as the average PC lifetime (5 years?).
do you have some data on that? some links? i've been trying to find something but so far unsuccessfully...

Well, the fact that PSUs have a continuous power rating should be good enough. If they can't supply that power continuously then it's false advertising.

nutball · Post by **nutball** » Mon Jun 11, 2007 5:47 am

mcoleg wrote:do you have some data on that? some links? i've been trying to find something but so far unsuccessfully...

TBH I think you should ask Mr "Guru" the exact same question. I mean if he's spouting so assertively, no doubt he has copious hard data to back up his claim. I mean a guy of all his experience, etc.

And "I use <blah>" is not hard data

mr. poopyhead · Post by **mr. poopyhead** » Mon Jun 11, 2007 7:21 am

uhm.... i dunno where people get these ideas from... about PSUs not being able to handle 50%+ loads for the long haul...

that's the most ridiculous statement i've ever heard.... who the heck would design a PSU that will fail if it runs at over 50% capacity for a year?? that's like saying that i should buy a ferrari for city driving because it has an easier time hitting 60km/h, while a toyota yaris would be running at a higher percentage of its capacity to reach the same speed.

continuum · Post by **continuum** » Mon Jun 11, 2007 9:27 am

mr. poopyhead wrote:uhm.... i dunno where people get these ideas from... about PSUs not being able to handle 50%+ loads for the long haul...

I can't figure that one out either. A quality PSU is rated to handle up to its continuous load, and unlike cheap pieces of junk, it will actually be able to do so. Period.

We haven't had thousands of systems with 12 7200rpm HDD's and dual
Xeons pass through extensive validation testing backed up by many years of customer use here on 460W PSU's if that wasn't true...

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 9:42 am

well, as an overclocker, i do get the idea of an overkill. also, as an overclocker, i'd like to choose my components carefully and with consideration.

so far it did not include the psu, apparently. a 500w unit would have been more than sufficient for the components and would have had a buffer to withstand the drop in efficiency from the additional heat. yet my last two psu buys were 700 and 750 watt.

good thing to question your assumptions, isn't it?

speaking of questions, i do ask any one who brings the topic of longevity up to show me some statistics/data/calculations/graphs/whatever (that included jonny as well). so far, nothing conclusive. i'll give it more time, i suppose.

does any one here has access to such data?

merlin · Post by **merlin** » Mon Jun 11, 2007 9:51 am

nutball wrote:
mcoleg wrote:do you have some data on that? some links? i've been trying to find something but so far unsuccessfully...
TBH I think you should ask Mr "Guru" the exact same question. I mean if he's spouting so assertively, no doubt he has copious hard data to back up his claim. I mean a guy of all his experience, etc.

And "I use <blah>" is not hard data

I'm very impressed by jonnyguru's articles and knowledge, however unless there's proof that running a low wattage power supply at higher than 50% of the rated capacity is any worse than running an insanely high wattage psu at the same amount of wattage(assuming all psu components are actually good quality), I'd like to believe there should be little difference in reliability.

Can anyone show through either testing or technical reasons that a high quality 380W rated psu at 250W is less reliable than a 700W psu at 250W.

Mr Evil · Post by **Mr Evil** » Mon Jun 11, 2007 10:08 am

Here's an extract from the ATX spec:

7.2. Mean Time Between Failures (MTBF)

the MTBF of the power supply can be calculated with the Part-Stress Analysis method of MIL-HDBK-217F using the quality factors listed in MIL-HDBK-217F. A target calculated MTBF of the power supply is greater than 100,000 hours under the following conditions:
* Full-rated load
* 120 VAC input
* Ground benign
* 25 C ambient

As you can see, power supplies must be able to run constantly at their full load - it's the law!

As for reliability data, here's a table from some random power supply manufacturers application note:

FPMH (Failures per million hours)
0Â°C 30Â°C 40Â°C 50Â°C 60Â°C 70Â°C
Temperature
Main Assembly 0.648 1.448 1.935 2.628 3.633 5.115
Channel 2 0.131 0.270 0.346 0.446 0.579 0.759
Channel 3 0.087 0.172 0.221 0.289 0.384 0.520
Channel 4 (â€˜xHâ€™ type) 0.053 0.137 0.196 0.286 0.425 0.638
Channel 4 (Linear) 0.001 0.020 0.039 0.074 0.136 0.240
Primary Option 0.054 0.307 0.541 0.958 1.705 3.028
Fan 2.396 1.941 3.067 4.845 7.656 12.096

The data is for temperature because that's what determines how long a PSU (and indeed most electronics) will last. It doesn't matter if you're drawing only 10% of the rating, if it's running at a high temperature it's not going to last long, whereas you can draw 100% load no problem if you keep it cool.

jaldridge6 · Post by **jaldridge6** » Mon Jun 11, 2007 10:26 am

I dont think any power supply would have a problem running at 50% load. It seems sorta silly to believe that a good power supply needs to run at a really low load in order to last a long time.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 10:31 am

Mr Evil - thanks a bunch man

nice linkage. i did measure the temperatures; i went even further and simulated higher heat loads (single post with pictures):

http://www.ocforums.com/showpost.php?p= ... tcount=115

the highest temperature i managed to record was 47.4C, heat load and power load combined. add some dust, what do you think, 50C sounds reasonable?

according to the pdf you linked, lambda psu will live with that load for 30 years. am i miscalculating something? that sounds like a lot of years...

this is a good answer to the question of longevity. what about partial degradation, how does that affect efficiency, any ideas? (sorry, i slept through Materials

)

Mr Evil · Post by **Mr Evil** » Mon Jun 11, 2007 11:07 am

mcoleg wrote:...according to the pdf you linked, lambda psu will live with that load for 30 years. am i miscalculating something? that sounds like a lot of years...

Be aware that MTBF is not the same as life expectancy. You'll be hard-pressed to actually find figures for life expectancy because it's not practical to test things for years or even decades.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 12:18 pm

got it, good point. reliability index, not life expectancy. still, the closest equivalent that can still be used in creating an argument, combining it with temperature data i've collected so far.

partial degradation of capacitors over a certain time-line - if any one has some data, please share

jaganath · Post by **jaganath** » Mon Jun 11, 2007 12:44 pm

partial degradation of capacitors over a certain time-line - if any one has some data, please share

http://en.wikipedia.org/wiki/Electrolytic_capacitor

electrolytic capacitors.

Since the electrolytes evaporate, design life is most often rated in hours at a set temperature. For example, typically as 2000 hours at 105 degrees Celsius (which is the highest working temperature). Design life doubles for each 10 degrees lower[1], reaching 15 years at 45 degrees.

most commonly there are two kinds, the cheaper wet electrolytics are rated 2000hrs@85C and the better ones 2000hrs@105C.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 1:28 pm

jaganath, thank you; i really should start using wiki instead of googling for formulas

i could probably, ruffly, say that the exhaust temperatures equal to the temperatures of the hottest component. that's not exactly correct but the only other way would be to take the psu apart and measure caps directly...

then, i could calculate the life expectancy of the caps.

as for MTBF, i need to think of a reliable formula. antec does not publish MTBF for ea series. on the other hand, seasonic does for it's own, similar psu's. 100000h at 25C. the temperatures are a bit unrealistic but what you gonna do.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 5:36 pm

a have another question.

in this post:

http://www.ocforums.com/showpost.php?p= ... tcount=182

there's a chart on capacitor aging. not sure i can make heads or tales of it but there's a chance it might explain something about an effect of a prolonged use on the efficiency.

any one has any ideas if this is relevant?

Mr Evil · Post by **Mr Evil** » Mon Jun 11, 2007 6:39 pm

mcoleg wrote:a have another question.

in this post:

http://www.ocforums.com/showpost.php?p= ... tcount=182

there's a chart on capacitor aging. not sure i can make heads or tales of it but there's a chance it might explain something about an effect of a prolonged use on the efficiency.

any one has any ideas if this is relevant?

I can't tell you how much effect capacitor aging has on efficiency (actually I can't seem to find any information on PSU efficiency change with time at all), but I can tell you that X7R and Y5V are classes of dielectric used in ceramic capacitors, which last much better than electrolytics. The graph shows the percentage change in capacitance with time. You can see that it drops as the capacitors age, by a lot for the cheap Y5V dielectric and by only a little for the better X7R.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 7:32 pm

http://www.jonnyguru.com/review_details ... page_num=3

these do look electrolytic though, don't they?

what i don't really get about that graph is how the time is measured (well, sort of get it but it seems that the time is indicated either for the restoration of the caps or for using them one hour a day?) and how the load is indicated. the load doesn't seem to be a consideration at all somehow.

Mr Evil · Post by **Mr Evil** » Mon Jun 11, 2007 8:07 pm

Big metal cans are electrolytic, yes. They're easy to spot, apart from being metal cans, they also have one lead marked as negative with a row of '-' symbols down one side. Non-electrolytic capacitors are never* marked like that. Also visible in the pictures are some ceramic disc capacitors which are disc shaped, obviously, and a few square-ish yellow and blue capacitors that are probably polyester film.

The explanation for that graph's axes was not clear. Do you know what the source was?

*Actually, rarely some wound foil capacitors have one lead marked, but that's so you can tell which is the outside of the winding, not which end should be negative.

mcoleg · Post by **mcoleg** » Mon Jun 11, 2007 9:07 pm

no clue what the source is. "company that manufacturers capacitors" was mentioned.

thanks for the explanations - i am so out of the loop... haven't looked at electronics since my school days; can't even recognize what's what... shame, that is

and sorry for the thousand questions; really do appreciate the input

continuum · Post by **continuum** » Tue Jun 12, 2007 11:50 am

mcoleg wrote:speaking of questions, i do ask any one who brings the topic of longevity up to show me some statistics/data/calculations/graphs/whatever (that included jonny as well). so far, nothing conclusive. i'll give it more time, i suppose.

I have data from our own production as well as customer data here, but it's proprietary. Suffice it to say that we run an ample amount PSUs near full load all the time and have been since production on certain products started more than 4 years ago, and we are seeing failure rates well within expectations.

Unfortunately I can't say very much more than that, but the failure rate is very. very low.

cmthomson · Post by **cmthomson** » Tue Jun 12, 2007 4:06 pm

In general, power supply lifetime is very strongly correlated with temperature. And of course load correlates with temperature, because power supplies are usually less than 80% efficient.

In general, power supplies fail because their electrolytic capacitors fail due to extended periods of high temperature.

Lots of generalities here, but (again) in general, a power supply will last indefinitely if it is kept below 70% of its rated load, and often will not survive to its rated MTBF if it is kept above 85% of its rated load. Of course this is also dependent on the fan speed, ambient temperature, altitude, humidity, etc, etc.

On the other hand, picking a supply that is rated at more than 2x the real load it will need to run is bad for (at least) two reasons: you'll be paying too much, and the power supply will be less efficient and waste more energy (most supplies are most efficient at 60-80% of rated load).

mcoleg · Post by **mcoleg** » Tue Jun 12, 2007 7:03 pm

thanks for the input guys.

cmthomson - this passage -"a power supply will last indefinitely if it is kept below 70% of its rated load, and often will not survive to its rated MTBF if it is kept above 85% of its rated load. Of course this is also dependent on the fan speed, ambient temperature, altitude, humidity, etc, etc. "- have you got any links on that matter?

collecting information and any further input on this will be appreciated.

continuum - i hope i am asking questions that are not against the rules here

- regular atx/server psus or some special ones? - constant load or variable?

disphenoidal · Post by **disphenoidal** » Tue Jun 12, 2007 8:26 pm

As a practical matter, overclockers and their ilk will likely buy a new power supply every couple of years for their Radeon/Geforce XXXX/ULTRA/OMG/WTF/FUBAR SLI/CROSSSSSSSFIRE upgrades anyway, so they really have no need to worry about longevity.