Article on PSU size & overkill; irresponsible journalism

jaganath · Post by **jaganath** » Tue Jun 12, 2007 10:44 pm

Radeon/Geforce XXXX/ULTRA/OMG/WTF/FUBAR SLI/CROSSSSSSSFIRE

mcoleg · Post by **mcoleg** » Fri Jun 15, 2007 8:18 pm

Mr Evil - he linked to this pdf:

http://www.rfeinc.com/FAQ%20&%20Tech%20 ... 0Aging.pdf

not the source but an interesting read.

jessekopelman · Post by **jessekopelman** » Fri Jun 15, 2007 11:55 pm

cmthomson wrote: 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).

So put these ideas together. First, for good longevity, do not run the PSU continuously at > 85%. Second, for good energy efficiency, run the PSU between 60-80% as much as possible. These suggestions do not contradict each other. Clearly, rather than using jonnyguru's 200% rule of thumb, we should be looking for a PSU rated at 125% our peak demand. For example, if you need 350W peak, get a 450W PSU. Seems simple enough to me.

One additional thing I'd like to point out is that for most people the computer is working hard enough to get to peak draw far less than 10% of the time. Even for hardcore gamers, I'd be surprised if it was more than 25% of the time. Even if you have a dedicated computer that just games, how many hours a year is it going to be in use? I bet less than the 2,000 used to calculate MTBF. Given that, I think you'd actually be safe to get a PSU that was rated for barely more than your peak draw and still expect many years of service.

mcoleg · Post by **mcoleg** » Tue Jun 19, 2007 4:36 am

i wish i knew...

trying to get lifespan numbers for a psu, no information on that. can't make an accurate prediction without it.

so, i have a couple of questions... again. sorry guys

1. where would one get a lifespan numbers for a psu?

2. MTBF; just as an example - sometimes it says simply 100,000 Hours @ 25C. at some places i saw 100,000 Hours @ 25C Ambient. so, which one is it? and what is the ambient - the temperature of the room or inside a case. now, this is the tricky one, seeing as the psu is a case inside a case, the ambient for it should be the computer case temperature; on the other hand it could be the outside temperature, i just don't know.

3. any idea why they are using MTBF instead of MTTF? i understand it's a multicomponent device but the weakest components are not really repairable and it's not like anyone actually repairs psu's nowadays.

4. please tell me if this statement is wrong - "modern psu is constanly drawing current, therefore there is no inrush current".

any ideas or leads would be appreciated, thanks guys

Mr Evil · Post by **Mr Evil** » Tue Jun 19, 2007 7:14 am

mcoleg wrote:1. where would one get a lifespan numbers for a psu?

Ask a manufacturer?

mcoleg wrote:2. MTBF; just as an example - sometimes it says simply 100,000 Hours @ 25C. at some places i saw 100,000 Hours @ 25C Ambient. so, which one is it? and what is the ambient - the temperature of the room or inside a case. now, this is the tricky one, seeing as the psu is a case inside a case, the ambient for it should be the computer case temperature; on the other hand it could be the outside temperature, i just don't know.

The temperature is always ambient temperature, which is the temperature of the air into the PSU.

mcoleg wrote:4. please tell me if this statement is wrong - "modern psu is constanly drawing current, therefore there is no inrush current".

There will be an inrush current from a cold start. From a warm start (i.e. without turning off the switch on the back of the PSU or the switch on the wall) then there will be much less. I would expect any modern PSU with active PFC to also include soft-start circuitry, which will mean a relatively small inrush current.

mcoleg · Post by **mcoleg** » Tue Jun 19, 2007 3:03 pm

Mr Evil, thank you.

manufacturers are not very talkative

the lack of information almost leads me to believe no one bothers to make those predictions... could that possibly be the case?

disphenoidal · Post by **disphenoidal** » Tue Jun 19, 2007 9:07 pm

mcoleg wrote: the lack of information almost leads me to believe no one bothers to make those predictions... could that possibly be the case?

I'm inclined to say that no one does. PC power supplies are consumer products, and like most consumer products, are designed to be essentially disposable. Made as cheap as possible and then blinged out. Nothing else matters.

Call me cynical. Being an engineer does that to you.

mcoleg · Post by **mcoleg** » Wed Jun 20, 2007 12:30 am

i do hope that's not true, there must be at least some starting point. you see, here's the thing - i am trying to prove that my initial calculations and following conclusions were erroneous.

have to do this myself since the best arguments against i've got so far basically sum up to a bucket-load of car analogies

i am almost ready for another set of tests that should prove or disprove the questions that were raised about survivability of smaller power supplies as opposed to large ones. however, even if i collect the data, i need a way to make predictions based on it. hard to do without certain basic information.

disphenoidal · Post by **disphenoidal** » Wed Jun 20, 2007 11:04 am

Is there anyone here who works with farms or clusters? I would imagine a company that uses computing farms would keep records of which failures were occurring and how often.

Post by **MikeC** » Wed Jun 20, 2007 6:53 pm

disphenoidal wrote:I'm inclined to say that no one does. PC power supplies are consumer products, and like most consumer products, are designed to be essentially disposable. Made as cheap as possible and then blinged out. Nothing else matters.

Not ALL power supplies. Retail packaged power supplies, yes. The rest of the PSUs for PCs fall into three broad categories: el cheapo for generic SIs, custom made for OEMs, and high quality for servers. With all of them, price is ALWAYS a factor.

An aside:

What we're seeing in the retail PSU marketplace is a brand-monetization (to coin an awkward phrase) of a sector (power supplies for DIYers) that was served mostly by the other three types of PSUs mentioned above. Prior to around 2000, the only "brand names" packaged for retail were perhaps Enermax, PCP&P, and Antec. Today, there must be some 30 brands offering retail packaged PSUs. And growing.

Until ALL of the DIY market is saturated with retail packaged PSUs, we will continue to see more brands offering PSUs.

mcoleg · Post by **mcoleg** » Wed Jun 20, 2007 7:40 pm

Mike, please elaborate. What's your point exactly?

Post by **MikeC** » Thu Jun 21, 2007 12:00 am

mcoleg wrote:Mike, please elaborate. What's your point exactly?

Not sure I had an exact point.

Just trying to add to the discussion, and more specifically to illuminate reasons for the preponderance of high power PSUs and why they are being marketed so aggressively. The main reason is that there's $$ to be made.

Also, that not all PSUs are made just on price; they have to survive a certain length of time under certain stress conditions for OEMs and server makers who won't be happy if there's a lot of warranty repairs.

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 4:31 am

i see...

i do have to disagree on one point - they all made "on price". there's no true "luxury" psu's left on the market ever since pcp&c went with cheaper oems.

as i look at the parts and the info i collect (and the info that is not available, which is more important) i can almost feel the souls of the engineers crying in agony

as they have to squish the designs into the given price bracket.

lots of unknowns and lots of fact juggling here. look at this page from seasonic and tell me that your questions are answered:

http://www.seasonicusa.com/s12plus.htm

and they are the good ones from the bunch.

jaganath · Post by **jaganath** » Thu Jun 21, 2007 4:35 am

remind me again what is the purpose of this research?

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 9:54 am

right now i am trying to prove that a smaller psu, like the one i tested, powering the system, as i tested, will NOT survive in a long run.

jessekopelman · Post by **jessekopelman** » Thu Jun 21, 2007 10:56 am

mcoleg wrote:right now i am trying to prove that a smaller psu, like the one i tested, powering the system, as i tested, will NOT survive in a long run.

Good luck

jonnyguru's position is ridiculous. He pulled his 50% number out of his ass. If he had said something like 70% or 80%, it would be more reasonable.

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 11:13 am

someone gotta do it

i do suspect though there's more to jonnyguru's opinion than he is saying...

mr. poopyhead · Post by **mr. poopyhead** » Thu Jun 21, 2007 2:35 pm

mcoleg wrote:right now i am trying to prove that a smaller psu, like the one i tested, powering the system, as i tested, will NOT survive in a long run.

i'm sure you can find TONS of anecdotal evidence on this board alone that would suggest that a PSU can be run long term at over 50% of its rated capacity (or 80% in your experiment)

perhaps your article conclusion should be changed to "i am trying to convince anyone to ditch their large PSUs in favor of smaller and more efficient units."

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 10:01 pm

anecdotal is not good enough; numbers, that what we are looking for

jessekopelman · Post by **jessekopelman** » Thu Jun 21, 2007 10:54 pm

mcoleg wrote:i do suspect though there's more to jonnyguru's opinion than he is saying...

Why? Even if there were, it would be the same anecdotal basis that you don't want. Anyway,why not just call/e-mail some PSU manufactures and ask them their view on running units at > 50% utilization?

disphenoidal · Post by **disphenoidal** » Thu Jun 21, 2007 11:45 pm

jessekopelman wrote: Anyway,why not just call/e-mail some PSU manufactures and ask them their view on running units at > 50% utilization?

I'm sure they would say you could run at 100% utilization indefinitely, whether or not they knew this was true. However, they would probably qualify this answer with a reminder about leaving headroom just in case, how they have a new 1.5kW model coming out soon that you should look into, and the importance of LED fans.

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 11:50 pm

jessekopelman - "Even if there were, it would be the same anecdotal basis that you don't want."

actually, it's based on some of the experiments i've done and the experiences i've had recently.

to put it simply - the loads don't kill power supplies, people do

hard to tell someone he needs only a 300-watter when you know he will not arrange for a proper airflow and never maintain it properly and in a year or so it will be packed with dust and running hotter than an oven.

"their view on running units at > 50% utilization" - thanks for the suggestion. i am collecting questions right now i want to ask. i figure, with half a dozen threads and the article, we've sold enough antecs and seasonics in the last month for them; they owe us. we can take our cut in information

.

mcoleg · Post by **mcoleg** » Thu Jun 21, 2007 11:51 pm

disphenoidal wrote:
jessekopelman wrote: Anyway,why not just call/e-mail some PSU manufactures and ask them their view on running units at > 50% utilization?

I'm sure they would say you could run at 100% utilization indefinitely, whether or not they knew this was true. However, they would probably qualify this answer with a reminder about leaving headroom just in case, how they have a new 1.5kW model coming out soon that you should look into, and the importance of LED fans.

oh, that's a good one

jaganath · Post by **jaganath** » Fri Jun 22, 2007 12:53 am

mcoleg wrote:anecdotal is not good enough; numbers, that what we are looking for

because of the way the components in a PSU age, extrapolations from the short high-temperature burn-in testing that manufacturers carry out might give you a reasonable handle on lifetimes; however I don't think they give out this information. The kind of long-term studies you want just don't exist in the field in any kind of organised way.

hard to tell someone he needs only a 300-watter when you know he will not arrange for a proper airflow and never maintain it properly and in a year or so it will be packed with dust and running hotter than an oven.

very good point. the conditions that a psu has to work in can vary hugely; it may be that a 600W operated in that condition (ie neglect) will fail after 2 years, whereas a quality 380W under the same load in a well-ventilated case might last 5 years or more. I still maintain that if the case has decent airflow (as per the ATX standard) and is cleaned (dusted) on a regular basis, any utilisation point between 10-80% will not reduce the design life of the PSU (which is what, 5 years?).

mr. poopyhead · Post by **mr. poopyhead** » Fri Jun 22, 2007 6:45 am

mcoleg wrote:anecdotal is not good enough; numbers, that what we are looking for

anecdotal evidence is a perfectly legitimate way of collecting data for statistics. not all studies are done by experimentation... and it seems that with the lack of experimental data out there, it may be your best option...

Mr Evil · Post by **Mr Evil** » Fri Jun 22, 2007 7:17 am

mr. poopyhead wrote:anecdotal evidence is a perfectly legitimate way of collecting data for statistics. not all studies are done by experimentation... and it seems that with the lack of experimental data out there, it may be your best option...

Anecdotes are interesting, and may help to direct the course of study, but drawing conclusions from anecdotal evidence is a very bad idea due to the fact that it's very far from unbiased, e.g. it's more likely that someone having a PSU fail early will post about it than someone who used one without issue for many years.

mr. poopyhead · Post by **mr. poopyhead** » Fri Jun 22, 2007 9:06 am

Mr Evil wrote:Anecdotes are interesting, and may help to direct the course of study, but drawing conclusions from anecdotal evidence is a very bad idea due to the fact that it's very far from unbiased, e.g. it's more likely that someone having a PSU fail early will post about it than someone who used one without issue for many years.

i suppose anecdotal evidence was the wrong term for what i mean, and what mcoleg is looking for... the data he is looking for does not neccessarily have to come from experimentation. i think if he gathered statistics on the lifespan of our PSUs, it would be pretty compelling evidence. i think in the absence of published data and a lack of practical method of testing (unless he wants to take years of data...), collecting statistical data may be good enough... for now.

jessekopelman · Post by **jessekopelman** » Fri Jun 22, 2007 2:57 pm

mcoleg wrote:actually, it's based on some of the experiments i've done and the experiences i've had recently.

You realize that, by definition, experiences you've had recently are anecdotal evidence, right? Same goes with your experimental results, until they are verified through repetition by some unbiased party.

mcoleg wrote:hard to tell someone he needs only a 300-watter when you know he will not arrange for a proper airflow and never maintain it properly and in a year or so it will be packed with dust and running hotter than an oven.

Why are you trying tell such a person anything? Everything is relative. Obviously, it would be stupid to expect to run a PSU at full load at ambient temperatures that are out of spec -- just as it would be stupid to run your PC submerged in a bathtub full of water. Manufacturers publish the temperature and humidity ranges that they consider proper for operation alongside information about wattage. Why would you be concerned about being oversafe on one thing (wattage) yet completely disregard the other (temp)? It's starting to sound like what you really want to do is establish some correlation between wattage headroom and the ability to run the PSU in out of scope conditions (e.g. if I run it at < 50% wattage maybe I can run it at +10C temperature).

mcoleg · Post by **mcoleg** » Fri Jun 22, 2007 5:06 pm

hehe, subjective relativism

jessekopelman - lets not go there. my experiences are factual to me, anecdotal to everyone else. the experiments, however, were in part replicated by a third party (check the article for the link.)

i am hardly doing it to prove a point; i am just curious.

mr. poopyhead - i will try to continue to gather statistics on the lifespan of psu's but it's a slow and unfruitful process.

Mr Evil is right - negative statistics have positive buoyancy.

jaganath, jessekopelman - i have outlined some tests that would correlate temperatures and load.

here's what i am thinking (after a brief rant

) . psu advertising is a con game; at least that's what it looks like to me after reading dozens of labels, sites, reviews. think for yourself:

- psu's (good ones, at least) are rated at 40C operating temperature. so far so good. MTBF, however, is calculated at 25C ambient (that is if the temperature is specified at all). i already pointed out that for psu, an ambient temperature will be the case temperature, not the room temperature. that is not very realistic.

-the real info about psu life span either hidden from the general public since it's not helpful to the sales or simply is not researched at all.

-any useful information like for example continues vs. peak power output is often glossed over either because of bad marketing or because (in case of low-quality units) it's not helpful to sales.

there's more; i am sure... the end result being - the information is difficult to gather and correlate.

so, i am thinking - m/b i am going about it the hard way? try something simple, i say.

back to a simple premise, then - psu's are killed because of the high temperature, not the loads.

and here's a simple experiment - two psu's of the same model with different power ratings. one is 380w - most people will think twice before using it with my system; second is 500w - practically everyone is in agreement that it's plenty even in a long run.

run loads, measure temperatures with both. if the temperatures are close to each other, 380w is enough for a prolonged use; if they are significantly apart, it's not.

what do you think?

mr. poopyhead · Post by **mr. poopyhead** » Fri Jun 22, 2007 10:07 pm

mcoleg wrote:and here's a simple experiment - two psu's of the same model with different power ratings. one is 380w - most people will think twice before using it with my system; second is 500w - practically everyone is in agreement that it's plenty even in a long run.

run loads, measure temperatures with both. if the temperatures are close to each other, 380w is enough for a prolonged use; if they are significantly apart, it's not.

correct me if i'm wrong, but assuming 2 PSUs (500W and 380W) have 80% efficiency at a given output (let's say 250W), then aren't they BOTH going to be pulling 312W from the wall, and giving off 63W as heat? i know this is an overly simplified model... but i think you get the point.

well you might say that the higher power PSU is designed for higher temps, but then you're really comparing 2 different PSU designs... i think the more accurate experiment would be to take PSUs from the same line (like the liberty series) and see if the temps differe significantly between low and high power models.

mcoleg wrote:mr. poopyhead - i will try to continue to gather statistics on the lifespan of psu's but it's a slow and unfruitful process.

a poll in the forums would do nicely. though you'd have to carefully select the criteria for it. do something like they do on storagereview.com... they have a database of service life for different hard disks... it's not exactly the most exact science and it's not repeatable experimentation... but the stats gathered are quite useful. of course, they are working under the assumption that people aren't lying about their hard disk life... why would anyone have reason to? unless you're some kind of seagate fanboy trying to inflate the stats.

i think people here are honest enough to give you some decent stats on this stuff. unless we're under the assumption that this forum is full of deranged lunatics who are clinging onto voodoo beliefs about a 300W PSU being good enough to run anything, while hiding the fact that they are actually burning out every few months.

but it really is good to see someone in the OC community trying to be objective about this... actually, i'm sure there are plenty of sensible people in the OC community, but most of the get drowned out by the fanatics... honestly, sometimes i wonder how they can ignore the FACT that a high-end system still only pulls less than 400W... i guess once we can prove that a PSU CAN be run at over 50% capacity long term, they'll have nothing left to justify their PSUs with... besides ego...

i wish you the best in your research...