A [H]ard Look at Power Supplies

[[Spectre]]

I don't know if this has been addressed as I haven't read all of the previous posts.

My concern is that all conclusions are being made off of a single test specimen. Statistics would say that no valid conclusions can be made from a single specimen as a PSU that either fails or survives may easily be a unit that happens to fall on one side or another of a standard bell curve (and thereby not represent "normal"). I know that testing 10 units is not practical, but I would be interested some comment on this aspect of testing. Thanks,

~PJ

Sample size is always part of the kicker on doing any review as you are limited by time, money, and space. If a unit fails we will attempt to aquire another and verify the results of of the first unit of course. If a unit passes it gets a thumbs up because the method of selection of the unit is random. If we suspect the unit is not random for any reason we will of course attempt to aquire another to test a random unit. In this manner it is kind of like QCing a batch in that you rely on the randomness of the sample.

 

[FrgMstr]

I don't know if this has been addressed as I haven't read all of the previous posts.

My concern is that all conclusions are being made off of a single test specimen. Statistics would say that no valid conclusions can be made from a single specimen as a PSU that either fails or survives may easily be a unit that happens to fall on one side or another of a standard bell curve (and thereby not represent "normal"). I know that testing 10 units is not practical, but I would be interested some comment on this aspect of testing.

That is why statisticians do not review hardware. This is also why we suggest you do not use HardOCP as your single point of reference on any hardware. If you have concerns with the number of samples we test negatively impacting the validity of our conclusions, I would suggest you put no merit into our testing and do not rely on HardOCP for any of your purchasing decisions.

 

[arklab]

One thing that kinda bothers me is the use of the Tripp-Lite LCR2400 Line Conditioner.
That seems to invaladate any results as far as "real world" use is concerned.

I really doubt that any average user will have such a thing, and how a PSU handles "real world" power line problems is critical to most of us.

I'd prefer you bombard the PSU with plasible current input problems, and see how it can (or can't) handle it.

Just my 2 cents.

 

[unhappy_mage]

One thing that kinda bothers me is the use of the Tripp-Lite LCR2400 Line Conditioner.
That seems to invaladate any results as far as "real world" use is concerned.

I really doubt that any average user will have such a thing, and how a PSU handles "real world" power line problems is critical to most of us.

I'd prefer you bombard the PSU with plasible current input problems, and see how it can (or can't) handle it.

Just my 2 cents.

Electricity in the testers' area might be of better or worse quality that the average [H] user. Using line conditioning levels the playing field for everyone. Everyone can buy a line conditioner, so it's a reasonable test field. If they cost $5k it'd be a bit unreasonable, but I think the [H] crew have this one just right.

 

[[Spectre]]

One thing that kinda bothers me is the use of the Tripp-Lite LCR2400 Line Conditioner.
That seems to invaladate any results as far as "real world" use is concerned.

I really doubt that any average user will have such a thing, and how a PSU handles "real world" power line problems is critical to most of us.

I'd prefer you bombard the PSU with plasible current input problems, and see how it can (or can't) handle it.

Just my 2 cents.

The line conditioner is neccesary in order to produce reproducible results from review to review. With it we are able to be sure of our input parameters and then vary them in realworld ways by using our variac in test 2 eliminating confounding factors.

 

[jonnyGURU]

One thing that kinda bothers me is the use of the Tripp-Lite LCR2400 Line Conditioner.
That seems to invaladate any results as far as "real world" use is concerned.

I really doubt that any average user will have such a thing, and how a PSU handles "real world" power line problems is critical to most of us.

I'd prefer you bombard the PSU with plasible current input problems, and see how it can (or can't) handle it.

Just my 2 cents.

Read the article.

Read the thread.

The line conditioner comes BEFORE the Variac that the power supply will be getting it's power from.

So.. ONE MORE TIME... the line conditioner is used so input voltage is consistant from day to day. The Variac allows the reviewer to adjust the input voltage to simulate a brown out.

 

[Todd_a]

The setup looks good. I like all the perameters. Instead of a lowered input voltage... I would rather see results for lower heat. Run duplicate tests at 30'C. Most cases good just do not get that hot. Newer cards vent their heat out the back and most good cases have multile 120mm fans. Then you can compare the results between 30'C and 45'C environments and see if there are differences in the ripple, failures, and efficiencies.

Other things like noise at different loads and temps is nice.

 

[pevo]

Maybe this has already been addressed, if so, my apologies.

First off, wow. The set up is extensive and I laud your efforts to put together a tough metering system for power supplies. The marketing hype in this area has really gotten out of hand.

To the point: Sustained load is an excellent point to observe ripple/noise on the delivered rails. But a much more potentially troublesome issue is transient loads. Like when your 4 disk raid array suddenly wakes up, or when your duel vid card setup jumps from desktop rendering to pushing massive amounts of polygons (although really thats probably a drawn out load -- over several seconds). Any ideal 'instantaneous' transition in load is effectively infinite in bandwidth and can drain power caps and float ground. So, I don't expect any system to handle it flawlessly; but a truly solid system should be able to minimize ground bounce and ripple during those transitions. Is anything set up to test that?

 

[jonnyGURU]

The setup looks good. I like all the perameters. Instead of a lowered input voltage... I would rather see results for lower heat. Run duplicate tests at 30'C. Most cases good just do not get that hot. Newer cards vent their heat out the back and most good cases have multile 120mm fans. Then you can compare the results between 30'C and 45'C environments and see if there are differences in the ripple, failures, and efficiencies.

I kind of agree with you there. When Paul said he was doing 45C, I was like "eh?"

If I have a PC with ambients higher than 40C, I'm freaking out.

Even a quad-father with a pair of 8800GTX cards is around 35C during 3DMark06 loops and Prime95 with just a pair of 120MM fans.

They made a thread where they asked, "what is temp is your PC running?" and most people said around 45C. But they're using motherboard monitoring to get that number. The machine sitting at my feet is running an average ambient temp of about 30C in a 24C room, but SpeedFan says it's 40C. The air coming out of the back of the PSU is actually only 32C. But who measures temperatures with an actual thermometer?

 

[[Spectre]]

The flipside is a unit can do what we ask at 45c without exhibiting any charactersitics out spec then at temperatures lower it will be above and beyond that. Like I said in the editorial my ambient summer temps get above 25c so the lowest I could test would 26c +1c assuming perfect conditions.

 

[jonnyGURU]

I don't completely disagree with 45C either. A power supply SHOULD be able to function at those temperatures, but at what load?

All power supplies have a de-rating curve and hardly any of them publish what it is.

And even when a PSU gives you "operating temperature range" they don't tell you at what load they can operate at that temperature.

 

[[Spectre]]

. A power supply SHOULD be able to function at those temperatures, but at what load?

We shall see......and that is why I have my fire extinguisher charged


And the derating will curve come into play a few times I would suspect.

 

[paulinus]

Great! I'll be glad to read some really good PSU review )
have fun testing!

 

[qdemn7]

One thing that kinda bothers me is the use of the Tripp-Lite LCR2400 Line Conditioner. That seems to invalidate any results as far as "real world" use is concerned. I really doubt that any average user will have such a thing, and how a PSU handles "real world" power line problems is critical to most of us. I'd prefer you bombard the PSU with plausible current input problems, and see how it can (or can't) handle it.

Frankly any one who spends $$$ on a PC these days and doesn't have at the very least a decent UPS with AVR is an utter fool, IMO. For the longest people were not willing to accept that a quality PSU was as much a vital piece of a performance based PC, this attitude has finally changed. But people still resist spending anything more than a $100 or so for a UPS, some people don't even think they should use one. Which is like saying "Oh I'll spend a $1000 on my quad core CPU, and $600 on my video card, and $400 on my memory and $300 on my mobo, but just WHAT do I need a UPS for?"

 

[FrgMstr]

The flipside is a unit can do what we ask at 45c without exhibiting any characteristics out spec then at temperatures lower it will be above and beyond that. Like I said in the editorial my ambient summer temps get above 25c so the lowest I could test would 26c +1c assuming perfect conditions.

I think 45C is no where near out of the question now days considering you may have 300 watts of video cards churning away plus all the other components. As we mentioned, we think that 45C is the UPPER END of the scale, but it is in no way unrealistic from my recent experiences with high-end hardware under stressful conditions. It is not uncommon for my computer room to get to the 90F/32C mark during the summer in Texas and I think there are more than a few people that share my experience.

So is 45C stressful and possibly hotter than many users will see? Yes it is. Is it uncommon and unrealistic? No. Are we wanting to put these PSUs into environments that are stressful and realistic? Yes we are.

I think the points made with HardOCP's 45C testing will be a beacon in the industry where it is not uncommon to sell you a PSU based on its operation at 25C. Bottom line is that it is going to ferret out the best PSUs you can buy for the money. If you don't run 45C, it is going to be all the better for you.

 

[LOWBOY]

This will be great to have another source of GOOD information on PSUs as most review sites reviews on PSU's are not worth reading. As I can atest to the fact as have seen
100's of posts where someone bought a piece of crap PSU based on reviews that were "worthless" such as the aspires and older ultra's that should be given to their worst enemy or used as a boat anchor if they weighed more. Look forward to the 1st review.

 

[Todd_a]

Well like I said 45'C is good for stressing the PSU, but realistically a case temp of 30'C is likely average. I was not sure if you were testing noise levels and efficiencies at the different loads. If you are than having the two temps to test at would be important. Maybe during the summer the higher temps on the PSU will cause the fans to run louder or even cause the PSU to fail under heavy load, but the rest of the year the thing might run wisper quiet even under load.

One thing you really need to test is a good range of common PSU. I rarely see Antec and Enermax tested even though these are what you are likely going to find in most computer retail stores. You really should have at least a few of those... Especially the new Antec 650w one... I'm looking hard at that one myself. A nice OCZ, one of the new BFG ones, Sparkle, Fortron, ThermalTake, PC Power and Cooling, and Cooler Master. Then throw in a few genarics for fun to see if you can smoke one. These are the one most people buy... including the enthusiast market. For the Antec and Enermax try some smaller (380-480w) and larger (650-1200w) ones and maybe an HE of fanless model for fun.

I know HardOCP has done some deals with NewEgg, so include prices and links to buy them there (NewEgg rocks) and maybe even see if you can get a 10% off link for the article for us happy readers. They might be happy to sponsor something like this. I have seen them sponsor other articles.

 

[Digitalcaveman]

I like my power supply, 1KW is what you need

 

[WarMace]

I like my power supply, 1KW is what you need

Fixed: I like my bank account, HardOCP review is what you need

 

[Gk22CoE]

With the power requirements of today a in-depth and thorough review like this is needed.

 

[[Spectre]]

Well like I said 45'C is good for stressing the PSU, but realistically a case temp of 30'C is likely average.

Not neccesarily.

I was not sure if you were testing noise levels and efficiencies at the different loads.

All of your questions were covered in the editorial unless I am missing something.

 

[quadnad]

So will we be able to read the first [H] PSU review tomorrow morning, or will that take a few more days? I'm eager

 

[FrgMstr]

Well like I said 45'C is good for stressing the PSU, but realistically a case temp of 30'C is likely average.

Son, you are going to have to go somewhere else for "average." I want to know what the sumbitch will do in Texas.

Sorry, I had to get a bit smartass on page 6 when this has already been explicitly responded to.

 

[FrgMstr]

What, Kyle isn't using 40 ppp? Still on page 3 over here.

So when will the first review be?

Na, I like my pages a bit shorter, make things easier to find for me when modding/adminning.

Going up now....

 

[Buck_Rogers25]

First up, good shtuff guys, [H] is generally the first place i check for hardware reviews, now even more so, and is probably single handedly responsible for turning me into an extreme cooling addict . Noticed you mentioned the not doing low end gfx reviews much any more kyle, shame as i often rely on them for server end systems i build for others, but i can see the financial reasons why.

There's been quite some banter around this topic regarding temperatures and the 45C testing, obviously there are parts of the world where this can be very common for case temps, just like to throw in another instance.

For cheaper server builds, and especially the type of quiet servers i tend to build which are largely passive these days, 40-45C case temp is not uncommon, and that's in the UK. Rackmounts can often be very hot if people dont have an AC in the same room or are running their own small quarter and half rack setups at home in tight spots, or in the instance of HTPC's aswell, often i'm buying Thermaltake 480W Purepower PSU's for these setups, just to guarantee a good clean source. Sometimes it's more than is needed but the PSU fans and an 80mm fan blowing over the HDs is really all i want to be hearing these days, even that is too much for my quiet tastes at times, and then there's the curve changes which have already been mentioned here. So yes in a performance setup i agree with what some are saying, but the same PSU's are'nt going to be used in well ventilated performance setups only, it's worth remembering that these reviews will be just as much use to people building home servers as performance rigs, dual redundant server psu's make sense on paper, but not when you run out of paper if you get me .

The point someone made about 100% load is a little silly, even in a graphics workstation rendering in max you only reach 80-85% load (based on personal MBM readings vs PSU rating), and thats chomping away at cpu, cpu and caching to HD like a demon at times. I cant really think of an instance outside of a paralell setup where you'd be using 100% load on the PSU for more than second or two, peak at startup for example. Although i'm sure there are a couple i'm not aware of.

I'm quite keen to see how some of the no fan silent units do in your tests aswell, if there's going to be a casualty i'd say it will be there. I've not been brave enough to risk buying one yet, just because of the reasons i've stated above, slightly too risky for me from a purely financial point of view.

I'd very much like to see say a Meanwell and a Manson put through their paces, different testing methodology being that neither is ATX, but it would be an interesting test, not that i'd ever expect you to take the time on such a fringe method of cooling. I have a Manson 60A 1-15v variable i was using for my peltier setup which has to be the most solid piece of equipment i've ever owned, but then it should be for the money it cost , £275 imported from hong kong hehe. Pelt setup has since been retired for passive WC with a huge 15x120mm radiator running with no fans, combination of electric bills and 3 delta 120s lead me away from pelts in the end, but i may go back some day. For now the PSU is being used as a bench PSU for electrical testing.

Pelts have always been a very fringe arena even for cooling fanatics like me due to their power requirements, but they did make me very aware of exactly what makes a good PSU, which lead to my choice of a bench PSU (the manson) as it's the only company i found giving me every piece of information i was looking for, it's such a shame that the same cant be said for ATX PSU manufacturers. What your doing here is commendable, expelling the marketing myths and providing reliable information at very high standards, something that other sites have gone part of the way to doing, but not quite fulfilled everything i personally seek from this kind of test.

Keep it up guys, really looking forward to the first couple of reviews. Especially after the movers killed my rig when we were moving in just over a week ago by ripping off a capacitor from the motherboard!!! Fried just about everything other than the PSU when i fired it up , and that's not going to be much use with new hardware . So new build time after i get the insurance through, and considering the quad core sli options i'm looking at for my new graphics workstation having a reliable PSU review is even more important, not to mention fortuetusly timed .

 

[sbuckler]

You have Seasonic confused with someone else. Seasonic typically uses mostly OST capacitors. Not always, and never Fuhjyyu's, but hardly the best. God? No way.



I can't speak for Spectre, but I'm 99% sure he's going to be opening the hoods of the power supplies and pointing out what brand caps are used, if the units have indy regulation, the quality of the PCB, etc.

I don't know that seasonic really is that great - that's just what the forums say (I have an antec which is *hated*, although apparently mine was made by seasonic so that's ok...). Now forums are populated by all sorts who claim brilliance but often don't have a clue. However they are the only places to go if you want to see more then the standard torture test. Check out http://www.badcaps.net/ (there's lots in there about psu's as well as motherboards).

 

[jonnyGURU]

I don't know that seasonic really is that great - that's just what the forums say (I have an antec which is *hated*, although apparently mine was made by seasonic so that's ok...). Now forums are populated by all sorts who claim brilliance but often don't have a clue.

You about summed it up right there. In forums, you're dealing with 2% fact, 98% B.S.

However they are the only places to go if you want to see more then the standard torture test. Check out http://www.badcaps.net/ (there's lots in there about psu's as well as motherboards).

LOL! I'm well aware of Badcaps, thanks.

 

[PedroDaGr8]

hahaha, I learned about badcaps from jonny. LMAO.

 

[LOWBOY]

You about summed it up right there. In forums, you're dealing with 2% fact, 98% B.S.
LMAO... how true..how true.

 

[GoodBoy]

So glad to hear [H] is doing this. =)

As far the temperature discussions, 45C seems fine to me. But a small graph of say 25C, 35C, 45C, and 55C with efficiencies, or reliable output power at each temp (for 3.3v 5v and 12v lines) would paint a useful picture of the PSU's abilities.

Not sure how time consuming it would be, but at the same time it might point out upcoming failure of a PSU if the output current is dropping rather noticably each step. Maybe you could do this as a "warm up" and do 1 hour at each temp, 100% loaded maybe. The rated wattage should be used as the load amount, since thats what they are advertising. Even though I would never want my PSU running even 80% loaded (I shoot for 50% myself)

Just a thought.

I just got an Enermax Galaxy 1000W dxx, and it seems really nice so far. Put my fluke DMM on the same 12v rail that my 7900gtx is attached, and saw no voltage fluctuation at all going from 2d desktop to playing doom3 (12.19 solid). I replaced an aging Antec Neopower 480, which I have actually liked (haven't peeked into it to see what caps its using). It has probably 2 years use on it, I'll send it down to you if you want to put it thru the rigors, as an aged PSU. I only replaced it because I know that the newer graphics cards are going to be power hungry, and like I said above I try to keep the PSU's rated output at about double what I really need.

I learned years ago that the brand didn't necessarily mean quality: saw an Antec fry at a LAN, had sleeve bearing fan and it had siezed. And that the wattage rating wasn't necessarily a trusty rating either: saw a generic (CompUSA. i know, i know) brand 600W psu's 5v line drop to 4.27v while 3d gaming. Nvidia's driver kept saying insufficient power. It was on an ASUS nForce2 mobo that stayed amazingly stable thru all that, only problem it caused was 3d gaming issues.

Im getting [H] just waiting on the first review. Let me know if you want to test this Antec Neopower 480.

GoodBoy

 

[Ozone77]

Paul, how did that first PSU test go?

I ask because I see a hole in the incubator theory (literally). I am not sure the incubator design accounted for an exhaust fan. Can the incubator maintain the 45C ambient temp when the PSU is cranking the fan(s) at max speed and all the air is exiting the incubator? Just curious.

Other than that, I think you guys put together a serious benchmark for testing PSUs. Good Job!

After that BFG review, I can see so many other test units being a let-down that I think the PSU forums are going to become quite lively

 

[[Spectre]]

Paul, how did that first PSU test go?

I ask because I see a hole in the incubator theory (literally). I am not sure the incubator design accounted for an exhaust fan. Can the incubator maintain the 45C ambient temp when the PSU is cranking the fan(s) at max speed and all the air is exiting the incubator? Just curious.

There is an exhaust vent in the back that the PSU's mount to using a bracket and a gasket.

So far the unit maintains the 45c +/-1c very well save for a small temp sag of around 4-6c as the fans first spin up...but the rebound time is about 2-3 minutes. Once it rebounds to 45c I start the clock.

 

[Virgulino Ferreira]

The ATX specification calls for power supplies to be rated at 25c (77F) [...]

Actually, this is what the ATX specification says:
(http://www.formfactors.org/developer/specs/PSU_DG_1_0.pdf)

6.1 - Temperature - RECOMMENDED
Operating ambient +10 °C to +50 °C (At full load, with a maximum temperature rate of change of 5 °C/10 minutes, but no more than 10 °C/hr.)

Congratulations, it's the best PSU test methodology I've seen so far.

As I understand, however, there is a fatal omission. As "pevo" pointed, stability under transient loads is utterly critical.

A PSU can pass the test and still be harmful to the computer.

No PSU test can be conclusive unless the following sections of the ATX specifications are exercised:

3.2.2 - Output Transient Response - REQUIRED
Table 8 summarizes the expected output transient step sizes for each output. The
transient load slew rate is [...]

3.2.7 - Capacitive Load - REQUIRED
The power supply should be able to power up and operate with the regulation limits
defined in Table 7, with the following capacitances simultaneously present on the DC
outputs. [...]
Output voltages should remain within the regulation limits of Table 7, for instantaneous
changes in load as specified in Table 8 and for the following conditions[...]

 

[jonnyGURU]

Actually, this is what the ATX specification says:
(http://www.formfactors.org/developer/specs/PSU_DG_1_0.pdf)

That's the range of the operating temperature. Not the rating temperature. Stating the operating temperature is merely stating "this is the range that this power supply can be used", which is 10 to 50C. It does not state that this it is the temperature at which the power supply has to put out it's maximum sustained power.

As I understand, however, there is a fatal omission. As "pevo" pointed, stability under transient loads is utterly critical.

A PSU can pass the test and still be harmful to the computer.

No PSU test can be conclusive unless the following sections of the ATX specifications are exercised:

I'm aware what the ATX12V specifications state for recommended load percentages, but putting a load on the 3.3V and 5V that are EACH 30% of what those rails are rated for an only 40% and 60% of what each of two +12V rails are rated for are grossly unrealistic.

There is very little using 3.3V and 5V these days. The ATX12V specification is antiquated in this respect. Note that they still suggest a separate rail for the CPU and a 240VA limitation. If we were still adhering to ATX12V specifications, we woudln't have power supplies with enough power to power any SLI machine with more than a pair on 7800 cards. The documentation does take into account for power supplies with more than two +12V rails either. Essentially, the whole almost-two-year-old specification only manages to confuse the user looking for a good power supply for their new gaming rig.

EDIT: And before anyone points out the June 2006 date on page one of that doc, I'll point out that the document linked is a compilation of all current power supply form factors. The latest ATX12V specification is 2.2 and it's dated March 2005.

 

[davidhammock200]

I'm aware what the ATX12V specifications state for recommended load percentages, but putting a load on the 3.3V and 5V that are EACH 30% of what those rails are rated for an only 40% and 60% of what each of two +12V rails are rated for are grossly unrealistic.

There is very little using 3.3V and 5V these days. The ATX12V specification is antiquated in this respect. Note that they still suggest a separate rail for the CPU and a 240VA limitation. If we were still adhering to ATX12V specifications, we woudln't have power supplies with enough power to power any SLI machine with more than a pair on 7800 cards. The documentation does take into account for power supplies with more than two +12V rails either. Essentially, the whole almost-two-year-old specification only manages to confuse the user looking for a good power supply for their new gaming rig.

EDIT: And before anyone points out the June 2006 date on page one of that doc, I'll point out that the document linked is a compilation of all current power supply form factors. The latest ATX12V specification is 2.2 and it's dated March 2005.

These published specs are so out of date that any PSU that was based solely on them would be useless for today's gaming rigs!

 

[Virgulino Ferreira]

Actually, this is what the ATX specification says:
6.1 - Temperature - RECOMMENDED
Operating ambient +10 °C to +50 °C (At full load, with a maximum temperature rate of change of 5 °C/10 minutes, but no more than 10 °C/hr.)

That's the range of the operating temperature. Not the rating temperature. Stating the operating temperature is merely stating "this is the range that this power supply can be used", which is 10 to 50C. It does not state that this it is the temperature at which the power supply has to put out it's maximum sustained power.

What about that "At full load" in the quoted specification? If the PSU is rated as 500W, and operates at full load at 50oC, doesn't that implies that the PSU is 500W rated at 50oC (or more)?

I'm aware what the ATX12V specifications state for recommended load percentages, but putting a load on the 3.3V and 5V that are EACH 30% of what those rails are rated for an only 40% and 60% of what each of two +12V rails are rated for are grossly unrealistic.

These percentages are the required transient step sizes that the PSU must be able to handle. They are not the recommended load percentages.

In other words, the specification is saying that, if the PSU has a 12Vcpu rail capable of 20A, it must be able to handle a load change from 5A to 17A (12A = 20A * 0.6), at a rate of 1A per microsecond. And that load change can happen 10000 times a second, up or down. All this while maintaining the required 5% voltage regulation.

If we were still adhering to ATX12V specifications, we woudln't have power supplies with enough power to power any SLI machine with more than a pair on 7800 cards. The documentation does take into account for power supplies with more than two +12V rails either. Essentially, the whole almost-two-year-old specification only manages to confuse the user looking for a good power supply for their new gaming rig.

Actually, the ATX specs recommends an EPS12V PSU for "Extreme Gaming & Media Entertainment PC", with 600W or more.

As for multiple 12V rails, I personally hate it. Unfortunately, the 240VA limit is a legal requirement for many PSUs.


Cordially,
Virgulino.

 

[[Spectre]]

What about that "At full load" in the quoted specification? If the PSU is rated as 500W, and operates at full load at 50oC, doesn't that implies that the PSU is 500W rated at 50oC (or more)?

Doesn't change the fact that it is the recommended operating temperature range not the required output temperature.

 

[davidhammock200]

What about that "At full load" in the quoted specification? If the PSU is rated as 500W, and operates at full load at 50oC, doesn't that implies that the PSU is 500W rated at 50oC (or more)?

Sadly no, it is just the temp range over which the PSU will operate.

Continous 100% load is probably rated for 25C, like most PSU's.

If you are selling a high end PSU rated for continous 100% load at 40C/50C, believe me you YELL about it!

 

[Virgulino Ferreira]

Sadly no, it is just the temp range over which the PSU will operate.

Continous 100% load is probably rated for 25C, like most PSU's.

Operate at full load.

Full load = continuous 100% load.

 

[davidhammock200]

2. DON'T BE MISLED BY EXAGGERATED WATTAGE CLAIMS
To properly compare power supplies, wattage claims must state the maximum ambient temperature for continuous, full-load operation. Unfortunately for the consumer, this information is usually withheld, opening the door for manufacturers to exaggerate their wattage claims. They do so by assuming an unrealistic ambient temp of only 25°C (77°F), even though the actual internal power supply temp is at least 40°C (104°F). Since the proper full-load rating is 15°C higher for home use and 25°C higher for industrial use, these power supplies produce 33%-50% less power than their advertised ratings. See the derating chart on the right.

http://www.pcpower.com/technology/myths/

If you won't believe us, maybe you will believe PC P&C.