I own the Thermaltake 700w. Thank goodness i came across this review. It helped with my options of what to buy.

Its honestly not unbearingly loud as TR put it. TR is used to be around top of the line products all the time, so compared to what ever they consider low sound must be deafening loud for us! People listen to their Music,TVs and games louder than 56 decibels. How can people pay attention to a fan and be so picky lol.

couldnt pass it up it was 129.99 with a 50 rebate from Frys.

I have a OCZ Gamexstream 600 W. It has been very reliable for the past year or so. It's a real champ. Runs a overclocked quad core plus 8800 GTX without even a twitter of complaint.

The only issue I have with it is that it is noisy, like they say here. The 120 mm fan is the culprit, with a rather noisy motor. As the PSU warms up, that fan speeds up, and it is easily the loudest fan in my system. Not because of high fan RPMs, but just because the fan's motor is clickety. Heh.

I've considered replacing the fan but obviously that results in zero warranty.

Good review. Further testing you could include:

- Would it be worth looking at efficiency at 25% load? That's practically idle, and probably the area a PSU spends most time in.

- The aforementioned swings in demand

- Make one go boom :-)

Excellent article, and The Beast is just impressive!
I got an HX620 last year without knowing it was so good. The thing is amazingly quiet and have never had a hiccup.

Thanks for this superb article Geoff and TR!

JonnyGuru and HardOCP blow this article out of the water.

You're right. I mean, the TR guys didn't even discuss the importance of the marketing graphics on the shipping boxes for the PSUs, nor breakdown the differences in the user manuals. I mean, how can you call the review a serious undertaking without examining the packaging and manual?!!?One!1

By far the most timely article in my case. Just one thing though, none of these supplies will mean squat when the light go out. Please follow up with an UPS article. Then retest of power supplies when on UPS power. Questions to test ac riple under simulated and or true sine wave.

FYI, these are also Seasonic-manufactured models:

Corsair 620HX
Antec EarthWatts 500W
Antec Neo HE 550W
Antec TruePower Trio 650W

The UL mark (E104405) indicates this.

Other Antec PSUs tend to be fairly dodgy, using poor quality capacitors and have a tendency to up and die rather suddenly without warning.

Seasonic easily makes the best power supplies in the industry. Fortron Source/Sparkle also makes rather good power supplies at very attractive price points.

Wow that article image looks....weird. like a bundle of snakes or something ..

Damage, etc...

Further tests I would like to see, and they have a lot of relevance to quality:

Dynamic load response.

For example when you go from low load (standby) to boot load (50% or more) in under 1mS.

Use a small scope onto the power supply rails (all of them if you can) and watch the outcome as the boot process unfolds. A well-designed PS will maintain regulation limits under this situation, but poor ones will dip below the limits temporarily and possibly then overshoot to try and compensate. This can be a serious issue of a board gets over-voltaged, or the board is run in an unstable region during boot.

Run on a Prescot board with a dual graphics-card setup for worst-case real-world test.

This is really a glaring omission from your tests, and is a very important design criteria when designing power supplies.

A recommended inexpensive scope would be the 32xx picoscopes:
http://www.saelig.com/miva/merchant.mvc?Screen=PROD&Product_Code=PS...
www.saelig.com

Which are perfect for reviews since you can capture the data from them very easily for insertion into charts. I use one at home, while not perfectly accurate, are very easy to use, and has great capture depth for doing these sorts of tasks.

And as mentioned before, source variance is an issue that could be tested as well, though much harder to accomplish. I do have at my disposal a LARGE variac that I think is rated for 2KW that I would give to you, all you need to do is pay shipping, since it weighs about 40lbs.

-LS

The Antec NEO HE does not work with Intel DG965WH motherboards. Don't know why, but the board won't boot up. It just cycles on and off.
The Earthwatts works just fine.

Oldtech.

And this is why i still like TR after all these years. Good solid reviews, graphs, a really nice job, and a community that supports the site. Really nice job there JBI. Of course, nothing is perfect, and there has come up some good points for the future. But then, TR is run by smart lads to i guess they have something up their sleeves for the next time around.

But the following isnt really bad things to try:
* Testing with a conditioned line.
* Variable ingoing voltages.

And i would also like to add the following. It would be great with an "inside" shot of the PSUs if possible, and also if there is available info on any OEM that it being sold rebranded, etc.

Last and probably the hargest is that i would love to see how far the PSU's go before they croak. Of course, the last one is mostly a wishfull thinking and i want it for kicks. In the old days i think there was some fortran source/sparkle PSUs that were rated 300W and croaked at 470W or something as absurd, but then, IIRC it was at Toms so it might not have been accurate.

Well, I've been a secret reader for quite sometime.

I like to read TR reviews.
Hope for the round power supply review,
other brand like Gigabyte, HEC, Enhance
or more should have a chance to shine. :-)

Cool review. Finally the nature of The Beast is revealed!

2¢ : Didn't the 2 640MB 8800 GTS' interfere/contaminate with your noise tests? If it were me, I would've just had a fanless 8600 GTS hooked up to the test rig. But that's just me.

Odd, I didn't notice this review until today.
edit: I love that testing unit, that thing is very cool.

"During testing and calibration, I also managed to destroy a ThermalTake TR2 PSU, when an incorrect switch setting caused me to accidentally exceed the PSU's combined maximum wattage rating. A popping sound, a flash of light (visible through the PSU's exhaust fan opening), a puff of smoke, and that was all she wrote. Beast indeed!"

I think we should get an honorable mention for Themaltake's PSU, based on the fact that you destroyed it. Poor little guy!

In these PSU 'roundups' there's never any Tagan supplies reviewed. I always thought they were quite a major player in the game. Am I wrong?

I wonder how efficient the average laptop power brick is compared to these powersupplies. Most laptops don't use all that much power (mine tops out at 70w full load afaik), so from an energy-conservationists' view, it'd be interesting to see..

edit: looks like regular laptop power bricks are mostly in the 80%+ territory as well

Thought: How bout testing efficiency at 25% too? I only ask since most computers just sit at idle most of the time, and seeing as this roundup of PSUs is all 500W and above, 50% is still at least 250W, while most systems will idle far below than, many not even half that much.

Maybe just do a minimum load test instead of a minimum percentage load and test efficiency there? Like test all PSUs at 125W maybe?

Just trying to be helpful.

Sort of off-topic:

A typical system for me has 1 upper-mid-range processor, a mid-to-high range video card, one or two hard drives, one optical drive, and at most one other expansion card besides the video. Also three or four input devices.

Currently, that's an E6300, GF 7900GS, 2GB of RAM, 1 Seagate HD, a DVD burner, joystick/throttle and gamepad. What's a reasonably-sized power supply for a non-hardcore-gamer like myself? Currently I'm on a 380 watt Antec power supply that was scavenged from an original Sonata case. If I wanted to, could I get by on e.g. a 250 watt p/s or even lower?

Power supplies are a critical component that are often overlooked. Otherwise, my Neo 550 has served me well for more then a year now, so props there.

I design VRMs and other power managment on server motherboards, and I just find it a bit hard to believe that an the test setup was burning 225W on idle? Assuming the AC/DC PSU was 80% efficient at idle, that means that somewhere around 190W were being burned downstream on the mobo somewhere...at idle?

As others have mentioned it is very important to test a power supply across line variations. As the input line decreases the input current will go up leading to possible thermal issues. As the input line increases the voltage across components will go up either changing the performance of certain components (MOSFETs and Caps) and/or derating their life.

If you want to get a good idea about lifetime of a power supply, the best thing to look at is the bulk capacitance on the output voltage rails. If they are using Aluminum Electrolytics, you can bet you'll get a shorter life than if their using Al Poly's. Some research into the different types of caps out there would take this review to the next level.

Most switching power controllers specify an initial accuracy for the output DC voltage, but this can drift with temperature...some controllers handle this better than others based on the configuration of the internal bandgap. Testing DC accuracy over temp is also necessary. If it varies more than +/- 10%, you've got a problem. Otherwise, most of the converters downstream are designed to operate within 10%, so outside that range is a "don't care."

We all burn a lot of energy with our computers....glad to know people are interested in getting the high efficiency units.

Thanks for the good article.

I very much appreciate this review, but you guys that are referring to it as "excellent" are not up to date with the other sites doing these. JonnyGuru and HardOCP blow this article out of the water. Anandtech is coming close but until AT starts to measure the power of quality they are just shooting blanks in my book.

I think trying to do 11 PSU's at once is just too big of an undertaking. There is an entire list of things that Geoff missed. Just go read a "real" PSU review that uses industry standard ATX PSU testing equipment. The Beast makes for a good photo op and a good story, but it is far from being a piece of equipment that is up to the testing criteria that is becoming standard in the industry. Doing power supply review OK is what they have done, but there have not even come close to excellent.

I like the attempt at noise measurements, but the data seems inaccurate at the loads stated. I also like the system build numbers, some lower level system builders need this and it is a welcomed sight.

My suggestion to the TR guys is to at least measure up to the competition. Half baked reviews are not the norm here, but that is what you have posted.

All the blind butt kissing is not good for TR. I will crawl back under my rock now as the TR fan boys will certainly come after me with a vengeance instead of reading this post as critical and spot on feedback that will hopefully make TR better.

Is it just me or are the result summary for the efficiency % switched? the 75% load results seem to be switched with the 100% load results...at least on the 2 or 3 I compared...

Awsome review totally in classic TR style. The only thing I could have hoped for was more PSU carnage from The Beast. Sadly, it looks like you decided to play nice. (110% Load test, pretty please? =P)

Also glad to see the PC P&C take home a gold medal.

Nice piece. We finally got to see the Beast. And kudos to just brew it! for his handiwork. A man of many skills. The Corsair HX is rather impressive.

This first page remark, though, makes me wonder what qualifies as a generic PSU:

Incidentally, we tried to get a generic 500W model to throw into the mix, but were thwarted by a general lack of availability. It seems generics don't typically aspire to output capacities of 500W or greater, and those that do don't cost much less than brand names with similar wattage ratings. This is a good thing for consumers, of course.

There is this ( http://shop1.outpost.com/product/4587977?site=sr:SEARCH:MAIN_RSLT_PG ) Ultra 500 watt model for $49.99. Not quite generic, but cheap enough. Back about 5 months ago lots of us got an Ultra 600 watter for ~$7 after $65 MIR, posted in the Hot Deals forum ( http://techreport.com/forums/viewtopic.php?t=50139 ), but that's cheating.

So that's what happens when you feed an old decade box steroids!
I also managed to destroy a ThermalTake TR2 PSU
I thought these power supplies had over-current protection? I guess burning up is one way to stop the excess current.

I note that the PC Power and Cooling supply runs a single 60 amp 12V single output distributed to multiple output wires with no intervening current limit per wire. This grossly violates Intel's informal recommendation of a maximum of 22 amps available per output. There is a very, very good safety reason for this recommendation. The potential for a catastrophic fire should a low-impedance path happen between any power-output and and the ground-return, either by a component-failure (e.g: capacitor) or circuit-board failure. I have been heavily involved in electronic equipment design engineering and reliability testing for many years and I have personally witnessed 2 cases of circuit-boards catching fire due to a failure of inner-layer isolation betwen a power-feedthrough and the encirclng ground-plane - diagnosed as whisker-growth due to (probably) moisture trapped during the multilayer fabrication process. (Similar damage to such feedthroughs can also occur when the feedthrough is the receptable for a component -- such as a capacitor or the pin of a power-supply connector -- and a careless replacement has occurred where the feedthrough has been overheated or physically damaged.) In the event of a power-feedthrough to encircling ground-plane conductive-path failure, all the available current (in this case +12V@ 60amps) will attempt to pour through that path raising the temperature until a fire ensues. A hole will be burnt in the board around the feedthrough until it is sufficiently large to involve all the power-planes in the multilayer board. The byproducts of the combustion are carbon and the hole will continue to burn and enlarge until the current-density through the carbon around the hole is sufficiently low for the temperature to fall below flame-point and the fire will extinguish. A modern circuit-board compliant with UL flame-retardant specs will not continue to burn after this point. However, there is no saying what might happen with other components inside the PC while this merry fire is going on. And if an unrestricted 60 amps is available for any output, instead of the Intel-recommendation of ~20 amps max current-limit per output over each of the 2, 3 or 4 outputs, the burn may go on 3 times as long before the originating source self-extinguishes. So who knows what else might catch fire ??

For safety, I **personally** do not recommend ever buying a PC power-supply with any output having much more current available than the absolute maximum current required of that output for that specific PC hardware configuration. BTW, for any given current, the +12V has the greatest available power ( watts = current x voltage...) - convertible to heat in the case of a failure. And a modern PC demands very high total power from the multiple +12V outputs --- a quite different scenario for PC built a few years ago, where 5 and 3.3v were the big power consumers. But, hey, it's a free country, and I am NOT in any way a representative of any official body responsible for public safety.

Man, how could I forget to give thanks where it is due: thanks JBI and nice friggin job!

Personally, I'm pretty darn impressed by the Neo HE. Nice performance at a nice price.

Excellent review, as always. That rig is impressive!

I always find something to be critical about, but it's hard here. The only comment I had was that standard deviations are great for AC ripple, but a max ripple (or maybe a 90 percentile?) might be more telling. Hard to do, though, especially as I imagine the ripple takes a minute to settle down after shifting to a a new load (especially when you flip one of those 8A switches, or bridge a few and flip that, heh). Or am I wrong? I've never hooked an ATX power supply up to an oscilloscope myself.

Great job, TR.

Awesome review guys. There is a distinct lack of proper PSU reviews IMO, and this is just awesome. I look forward to more of these.