A [H]ard Look at Power Supplies

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,602
A [H]ard Look at Power Supplies - HardOCP’s [H] Enthusiast finally spins up to correctly cover computer power supplies. We show you how we will be doing it, what equipment we will be using and the methods to our madness.

Power supplies and the requirements placed upon them have greatly changed in the last few years. PSUs are now a “make or break” element in a high end system build that cannot afford to be skimped on. And when you buy a power supply, don’t you want to know that it can provide your system with quality power at the rating printed on the box while it is inside your case? That is exactly what HardOCP wants to know too and that is exactly the information we hope to be providing to.
Click to expand...
 
Last edited:
Excellent. I'm glad you guys are stepping into the ring. There's just not that many sites I know of doing this level of PSU testing. I'm looking forward to seeing what you guys put out.
 
Kyle, I am extremely impressed at the level of quality you have brought to the process. I especially like the use of the line conditioner and the incubator. I think the incubator will be the Achilles Heel of many PSUs.
 
There's something about the Waveform Monitor output of the SM-8800 you're going to want to point out to readers BEFORE they start flaming you for having "unrealistic ripple and noise figures." *cough*Enermax*cough* *cough*OCZ*cough*

The ATX12V specification states that ripple and noise are measured at the connectors with a 0.1uF ceramic capacitor and a 10uF electrolytic capacitor to simulate a system load. Unfortunately, if you don't point that out to people first, they're going to assume this measure is not taken and flog you for reporting ripple and noise numbers that may also reflect RF pickup from the BNC connections, etc.

FYI: These prerequisite caps are already installed behind the panel of the SunMoon tester, soldered to the inside of the BNC the scope leads screw onto.

The SM-8800 does not differentiate ripple from noise, HOWEVER the ATX12V specification dictates that NEITHER should exceed 120mV on the +12V and -12V and 50mV on the +3.3V, +5V and +5VSB. So, any waveform that exceeds these parameters whether in the form of ripple or noise should be considered "failed" regardless.
 
jonnyGURU said:
There's something about the Waveform Monitor output of the SM-8800 you're going to want to point out to readers BEFORE they start flaming you for having "unrealistic ripple and noise figures." *cough*Enermax*cough* *cough*OCZ*cough*

The ATX12V specification states that ripple and noise are measured at the connectors with a 0.1uF ceramic capacitor and a 10uF electrolytic capacitor to simulate a system load. Unfortunately, if you don't point that out to people first, they're going to assume this measure is not taken and flog you for reporting ripple and noise numbers that may also reflect RF pickup from the BNC connections, etc.

FYI: These prerequisite caps are already installed behind the panel of the SunMoon tester, soldered to the inside of the BNC the scope leads screw onto.

The SM-8800 does not differentiate ripple from noise, HOWEVER the ATX12V specification dictates that NEITHER should exceed 120mV on the +12V and -12V and 50mV on the +3.3V, +5V and +5VSB. So, any waveform that exceeds these parameters whether in the form of ripple or noise should be considered "failed" regardless.
Click to expand...

Thanks for the heads up on this jonny! Any feedback you have is very much appreciated.
 
Im planning on getting one of those modular Silverstone PSU's. Thinking about the 750 watt one. So a review on one of these would be nice:D Plus I hear Silverstone makes some nice PSU's anyway.
 
great guys... but I reall do hope I wont read about power supply reviews of 750W or over. Most people still buy 500W or lower PSU units (cause frankly how many people need more?)

keep up the good work
 
germanjulian said:
great guys... but I reall do hope I wont read about power supply reviews of 750W or over. Most people still buy 500W or lower PSU units (cause frankly how many people need more?)

keep up the good work
Click to expand...

Well.. this IS [H]ard|OCP and you can't really run a stable quad core with dual 8800GTX and more than a couple hard drive stable with anything less than a 700W. Dual R600 needs even more than that.

That said, I know what you mean. My wife's machine is an Athlon64 with a 6600GT and one hard drive and runs 24/7 with a mere Ultra V-Series 400W. But I'm not really concerned that it's not SLI ready or may not hold up if temps inside the case exceed 40C. :D ;)

I think for "Joe Average User" you should just stick with brands you know, good support, good warranty and if the vendor has a product feedback feature, see what others have to say about the unit. But if I just shelled out $400 a pop for video cards, I BETTER (hint hint) be looking to spend at least a buck and a half on a power supply and it better do what it's supposed to do.
 
Sound great Kyle, been looking for some serious PSU tests where they test the PSU under "normal" conditions ie. the way we users use them.

These PSU's sure needs to be tested at higher ambient temps (45°C or even higher) and be pushed to the limit with full load on all volt rails.
This is getting to be more and more important lately as our computers draw more power and when the chips get down to 65-45nm stable volt is very important.

Looking forward to some good reading here.:)
 
The only thing I see, is the assumption that a computer will max out a power supply for continuous periods of time. While reasonable, computers really don't take as much power as most people like to think: this SPCR article lists the actual power consumption of some of the more power-consuming components. Actual power draw is only slightly above 200w. In my case, a highly oc'd opteron, 5 hard drives, and a 7800gt, draws just over 100w full load. Not a particularly huge stress.

Just something to look out for.
 
Finally someone who will expose those unscrupulous companies that sell "500 watt" (or other) PSUs that they tested at 25C, when in reality the PSU puts out maybe 300 watts in a typical environment. PC Power and Cooling tests their PSUs (at least the server ones) at 50C. (not saying you should adjust your temps, only pointing out that honest testing involves honest temps, not hypothetical 25C case temps that DO NOT exist in high-end enthusiast systems for the most part)
 
Normally I just read Silent PC Review for PC power supply reviews. I think the [H] can complement SPCR by giving the viewpoint of the enthusiast/power users/overclocker crowd in addition to what we already have from the quiet/efficiency crowd.
 
germanjulian said:
... people still buy 500W or lower PSU units (cause frankly how many people need more?)
Click to expand...
If time allows, I'd really love to see some 'low power' PSUs as well. Much to my surprise my 375W Enermax Noistaker PSU is still going strong, even when running an X2 with a 7900 GT. So it'd be very nice to see what "low-power, low-cost, low-end" PSUs can still run nowadays.
 
hello,

on the strength of this article alone, i signed up for the forums.

this is an area of product testing that sorely needs the degree of rigor your proposed methodology supports.

as an amatuer system builder who does projects for others i have been arguing the point that the psu is one of the most important components in a system with every box i put together.

in the recent past we have seen interviews with Doug Dobsom of PC Power & Cooling (my psu of choice) make great points concerning the lack of standards in the psu industry and what is wrong with modular psu's. and then with Ryan Petersen of OCZ Technology which builds modular psu's agreeing with Dobson about modular designs.

its time for rational, rigorous, independent testing and it looks like [H} has got it dialied.
 
Nice to see another site doing some real power supply reviews as it's fairly rare. The conditions are fairly [H]arsh, but I'd expect nothing less!

On a similiar note, please consider adding actual power usage testing of components that are reviewed. Nothing in-depth just an actual max power draw, and what rail it's being drawn from, and ideally a max power draw when overclocked. Motherboards, video cards, hard drives, physics cards, etc. Combining some Hard numbers on power draw with Hard numbers on what power supplies can put out would be an amazing tool for enthusiasts. Yes I am greedy and want it all. :D
 
Bbq said:
The only thing I see, is the assumption that a computer will max out a power supply for continuous periods of time.
Click to expand...

Where exactly did you see us state that or infer to it?

There is the larger portion of a paragraph carved out advising against it. Now that said, should you have a quality unit that you will never press over the 50% mark, well then, you likely have no reason to read PSU reviews IMO.
 
I'm very glad to see this happening, and even happier to see you chiming in jonny...as I'm sure your level of PSU expertise will be much appreciated here at [H] just as it is on your own site/forum.
 
You know, I was thinking over the past few days why I haven't seen any power supply reviews from [H] because, in my opinion, they are at least methodical enough to undertake such a task to provide reliable results. Reviews here, indeed, would supply a compliment to SPCR's reviews.

Speaking of which, since the few sources I go to for reliable PSU reviews sometimes provide what I am about to request, would it be possible for [H] reviews to state somewhere who manufactured the test subjects (i.e.: Corsair power supplies come from Seasonic, Enermax makes their own, etc.), or even go so far as to test power supplies that originate from the same manufacturer? Granted, anyone who looks at a UL number on the side of their power supply could easily go to the UL website, type in the number, and find out who is really the manufacturer, but, while looking for a new power supply, a consumer might be considering two power supplies that are actually one and the same without knowing it. If a few power supplies are based off of the same manufacturer and design, similar power supplies could be tested simultaneously to see if there really is any differentiation. In the instance of Corsair HX series and the Seasonic M12 series, the only feature difference is the latter has a 60mm fan that compliments the 120mm fan they both share, but they're both modular and their internals are exactly the same so a Corsair HX620 VS Seasonic M12 600 article would make sense.

Just a thought.
 
Talz said:
Nice to see another site doing some real power supply reviews as it's fairly rare. The conditions are fairly [H]arsh, but I'd expect nothing less!
Click to expand...

Well the equipment is not cheap at all, but finding the right person has been the real challenge. Paul looks to be our man and has some great support inside the PSU community, and he has a passion for the subject matter. Getting all those qualities to line up is not easy, I wished we could have done this two years ago...and I was trying.

Also, an email asked about review formats. Likely we will start off with single units, then scale to roundups of multiple units. However if we keep our testing method standard, we should be able to compare across units even if done individually. :)

I am really excited about all of this. BFGTech stepped up for our first review of their 1KW unit. All of the testing is done, we are just left with some formatting for it.
 
Cool! Will only synthetic loads be used for these tests, or is testing big-scale rigs in the works as well? Might end up with a few RMAs then, but it's one more thing [H] could do differently from the non-crappy PSU review sites. Certainly not essential, but it'd be neat.
 
hell yes, [H] expands its repertoire!

I really look forward to seeing PSU reviews with the same fanatical attention to detail as the video card and system reviews. [H] has done a great job so far, keep it up!
 
yojimbo said:
Speaking of which, since the few sources I go to for reliable PSU reviews seem to provide what I am about to request, would it be possible for [H] reviews to state somewhere who manufactured the test subjects (i.e.: Corsair power supplies come from Seasonic, Enermax makes their own, etc.)?
Click to expand...

I dont see why we should not provide that information considering it is far from a "secret" given the public nature of UL listings. I know this is something that Paul seems to already have a great deal of knowledge about currently so it would make sense for him to bring that to the table.

And the fact is, if you can buy the same PSU for $X cheaper, more power to you....pun intended.
 
unhappy_mage said:
Cool! Will only synthetic loads be used for these tests, or is testing big-scale rigs in the works as well? Might end up with a few RMAs then, but it's one more thing [H] could do differently from the non-crappy PSU review sites. Certainly not essential, but it'd be neat.
Click to expand...

I have looked at bringing PSUs that "[H] Level" of real-world testing, but this is one area that is just falls on its face in. We simply do not have the resources in both terms of time and money to do it that way.

Keep in mind that the SM-8800 is specifically designed to load test ATX PSUs. Sun Moon quoted me $5000 for the machine we are using now. (We got it a bit cheaper.) But the point I am wanting to make here is that the SM-8800 is a very sophisticated power supply tester and I think it will do a great job.

Now when it comes to longevity, it is one of those areas that is not easy to pinpoint without running months or years of tests. This is one of the reasons we have set out testing standards very high as we think doing so will help us pinpoint the the best units in terms of durability and longevity as well.
 
Nice to see [H] finally doing PSU tests/reviews. I know for years that you've always been kind of tongue in cheek about other sites reviews of PSUs. But I never really understood why [H] wasnt doing them. Now I know and am looking forward to the first ones.
 
showing real work power draw is VERY usefull for people these days, and really does shatter the marketing image that you need an 80)W+ PSU for a C2D gaming system no matter what. I agree that it could get expensive/time extensive setting up a real machine for each test, but a single article devoted to showing real world power draw of a few Tiers of systems would be nice.

Keep in mind though that the focus of all this is to show capabilities of the power supplies themselves, not necessarily the draw of the systems they're running. I also understand that people will have to know how much power their system might actually be drawing if they're gonna line it up with a synthetic load on a PSU during a review, but a single article to show a couple/few AM2 as well as a couple/few C2D systems real world draw would be handy.
 
3) Torture Test

a. ...and a load approximately equal to 80% of rated continuous output is applied to the power supply by the Sunmoon SM-8800 for a period of 8 hours
Click to expand...

If a manufacturer says that a power supply can deliver 500 watts continuously, how can one say that running it at 400 watts for 8 hours is a true test of its power rating, much less a torture test?

Wouldn't that be like testing a 2 GHz CPU by underclocking it to 1.6 GHz and running it for 8 hours? Nobody would consider that a legitimate CPU test. On the contrary, torture testing a CPU means overclocking the CPU and testing it, usually for longer than 8 hours.

If the problem is that most PSU's can't deliver what they promise, isn't it your job to identify them instead of skirting the issue so as to avoid embarassing the PSU manufacturers?

It seems to me that you should test PSU's at 80%, 90% and 100% of their continuous rated capacity, and let the sparks fly where they may. That way, the PSU's that are truthfully labelled are not penalized.

Fuzzy
 
cornelious0_0 said:
showing real work power draw is VERY usefull for people these days, and really does shatter the marketing image that you need an 80)W+ PSU for a C2D gaming system no matter what.
Click to expand...
True, but then a quality 1kW unit driving a 200 watt load is going to run a lot cleaner, cooler and quieter than a 500 watt unit driving the same load, assuming we're talking equal build quality.
 
Bada-Bing! said:
Normally I just read Silent PC Review for PC power supply reviews. I think the [H] can complement SPCR by giving the viewpoint of the enthusiast/power users/overclocker crowd in addition to what we already have from the quiet/efficiency crowd.
Click to expand...

Why does [H] need to compliment another site?
I think [H] can stand alone on this!!
Also SPCR is NOT the definitive voice on PSU testing at present I honestly believe that Honor would fall into JonnyGURU.com site!!
 
This is great. I love Jonny's reviews, but he can only do so many of them, and it will be good to see another reviewer doing a great job.
 
I see nothing to measure noise!

I realize that many enthusiasts may not care how loud their PC is, but I personally like to avoid having my rig sound like a leaf blower.

Comparing the "rated" dba and actual, as well as subjective listening is something I think should be included in any PSU review.

Just my 2 cents!
 
Awesome!!

I now have two excellent sites to check out information on power supplies. Both [H] and Jonnyguru will be my hookup on info. Are we allowed to make suggestions on what companies we want to see reviewed?
 
OCC_Yoda said:
Why does [H] need to compliment another site?
I think [H] can stand alone on this!!
Also SPCR is NOT the definitive voice on PSU testing at present I honestly believe that Honor would fall into JonnyGURU.com site!!
Click to expand...

I agree 100%...but it'll be nice to have another resource...especially with jonny lurking [H] in that regard. :D
 
What about measuring how well these psus hold up against dirty power? The line conditioner pretty much guarantees good, clean, stable AC in.
 
Zoomer said:
What about measuring how well these psus hold up against dirty power? The line conditioner pretty much guarantees good, clean, stable AC in.
Click to expand...

Doing it with the conditioned power puts all the power supplies on a level playing field. Without using a line conditioner, one power supply might do better than another one simply because someone is riding the mechanical bull in the next room while the tests are running.:D
 
You must log in or register to reply here.
Back
Top