A Seasonic 380W or 430W would've been a nice addition to the round up. And if you really like the Corsair VX450 but want a second PCIe line, the VX550 doesn't cost much more and adds a 6/8 pin PCIe connector (note: I believe they use different OEM's though, Seasonic for the VX450 and CWT for the VX550).

Sigh. No mention of internals, components, build quality, overload capacity, cross load votlages, fanspeed profiles? -_-. What about elevated temperature testing?

Nice article.

I just have a few comments on the efficiency testing. First, it might be good to mention what 80+ certification means: it's 80+ (or higher for higher certifications) from 20% to 80% load. You could also link to www.80plus.org

As for the actual efficiency results...I've got to question them. This is the first PSU review I've ever seen in which the efficiency curve is not a 'hump' with lower efficiency at very low and very high loads than in the middle. Also 90%+ efficient is virtually unheard of in these types of PSUs although there are a few brand new topologies that get up to near 90 with 240V input, those topologies are not used in these PSUs. I would suggest reexamining the test methodology for efficiency.

Good to see a Basiq included, since I just bought one of those at OfficeMax back in October for a second-hand build that stalled abruptly when the Seasonic SuperTornado 400W decided that it had just gotten too old. The Basiq seemed a little on the light side, so it's good to see it do well in your tests.

Couple points:

1) In the future, would you consider putting an actual digital thermometer probe somewhere in the upper part of the chassis and see what the ambient chassis temperatures look like independent of the hardware monitor?

2) Regarding this statement on p3:

"We generally like to see PSUs bias their output toward the 12V line—that's what processors and graphics cards tend to hit the hardest—so the SolyTech's preference for 3.3 and 5V power is a little puzzling."

http://techreport.com/articles.x/16073/3

Puzzling in what way? SolyTech appears to have simply built both rails equally using middle-grade parts in order to get their 350W rating, which allows them to stock just one set of switching transistor pairs and Schottky rectifiers. It's a garden-variety cost cutting measure, whereas the other manufacturers would have been able to reduce the requirements of the parts for one rail balanced against the added expense of stocking two part sets, and then spend extra to get a high-current 12V rail.

In short, SolyTech probably saved a couple shiny nickles per unit and produced an inferior quality unit without technically lying about the specs.

Cool read. Thanks. My only nit to pick is that you say that you 'singled out' the Corsair, even though it was a 2-way tie. Hard to be tied and yet single.

was supposed to be reply to #72...

The latest 80+ draft testing spec is available here:

http://www.80plus.org/documents/Server_Power_Supply_Efficiency_Test...

One noteable difference between the 80+ methodology, and TR's methodology, is that the 80+ method requires using very precise power meters on both the input AND output sides of the supply to verify what is actually flowing through the conductors. TR's output power measurement involved watching the voltage while assuming the output current could be derived from Ohm's law.

Here's a speculation: How hot do those resistor banks actually get during operation? Each of them will demonstrate negative positive temperature coefficient -- meaning, as they heat up, their resistance will increase, and there will be slightly less current flowing through the load than expected. The current flow to the input side will reduce slightly in proportion.

Has TR confirmed that the change is negligible, and the expected output power is accurate? If not, then efficiency calculation could be artificially inflated by some nontrivial amount. At lower loading values, like the 25% or 50% cases, the proportion of error would be large relative to the net loading, which might explain why the TR efficiency curves seem to have a preganancy figure (very high efficiency at the lowest loading values).

Edit: Technical correction.

Whenever someone suggests buying a generic PSU to save a meager 10-15 bucks, I remind them about the cost of replacing and show them the vid of the spectacular failures of generic PSUs Corsair made.

That baiq was a damned loud unit it seems. I wonder if it was a slightly bad fan?

I wish TechReport had included a 400 watt silverstone Strider in the test group, Newegg.ca was selling them for $27-$37 per unit limit 5 per order and NCIX.com is matching that but in unlimited quantities.

while their were good performing powersupply`s in this test group the prices I saw were terrible and $72 for the corsair even with the 5 year warranty would go ignored.

while it may sound cheap the price for a reasonable build dual core, 2gb ram, DVDRW, 320gb or 500gb hdd, case and powersupply is closer to $300 than it is to $350.... at $72 for a 400 watt psu it`s simply not going to happen and anything higher will go ignored.

I`ve sold the first 4 silverstone Strider 400 watts within hours of arrival and have another 10 coming it would have been interesting to see how they faired.

I hope and believe they would have done well.

I'd like to know what these PSU's use when the computer is still attached to it but turned off... In my experience the wattage varies quite a bit depending on the psu you are using... i have a couple of PSUs (from antec to fortron to recovered PSUs from dell desktops) and that wattage is varrying from 10W to 30W with the computer being turned off! A powersupply that *ABUSES* 30W of power while it's assumed to do nothing at all pretty much kills all 80+ certification out there...
(just for you to know: the dell desktop part is the most efficient ;) )

Thanks for the article.
Another aspect of interest is the quality of filtering on both mains input-rf filtering (keeping the hash in) and output-keeping the DC clean(low ripple on full load-indicating probable lifespan of supply.
A rough indication of rf filtering would be operating a sensitive AM radio nearby, and output filtering-scope the rails at full load and, say, 5% beyond.
Cheers.
Pedro

No sacrifices for the beast eh?

Just Brew it nice job on The Beast! I think the innards look sexy. I always wished I had gotten more into that kind of stuff. Looks very professional.

One small note:

In your last system guide (Christmas) you recommend the Antec NeoPower 500W Power supply and an Antec P182 for the The Sweeter Spot system. The 8-pin Cable for the EATX12V isn't quite long enough to be fed behind the motherboard, which was disappointing to find out... The system I built has that one lone cable over the motherboard...

It looks like Antec dramatically reduces cable length for the lower wattage-rated models, (as this review certainly proves) which can certainly impact system builds...

That was a great article and a very prudent choice of PSUs. I am a little surprised an Earthwatts has not made it to the bench yet. At $80, it is right in line with the segment and offers 500 watts of 80+ power.

This is the most relevant grouping of PSUs ( to me ) to date.

Thanks.

Just my 2 cents - Obviously not having sufficiently lengthy cables on a power supply blows, but at the same time theres those who like small cases & don't see any value to longer cables. I definately think documenting the cable length is a good idea but don't think that 'longer is better' always applies.

Great article as always.

I'm kind of surprised that the piece of crap no-name 350w power supply was able to run an SLI system under load. Then again, it will probably be dead within the month.

Great tip on the FSP store. http://www.shopfsp.com/clearance_center.html
With free shipping some of those deals looked very sweet.

Wow, just brew it sure does know his stuff - that article on the beast is complex.
I suppose he makes some good money in some kind of clever person job, I just get to reap the rewards by knowing what's best for my money :)

On page 3, you say: "The math is easy enough thanks to Ohm's Law, which allows us to calculate wattage given voltage and amperage." However the power formula (P=IV) is not Ohm's Law (V=IR or permutations thereof). It doesn't really have a name of its own, but it's closely related to Joule's Law.

Otherwise, nice article.

This may sound goofy, but could you post the weights of the PSUs in the chart? Directron had an interesting analysis a couple years ago showing a direct correlation between PSU weight and quality.
http://www.directron.com/psu.html
It would be interesting to see if this correlation remains true.

not sure if this was mentioned though it doesn't matter, but the 'system temperature- load- hard drive' yellow and orange appear to be reversed

I just replaced a cheap 450W PSU with the Corsair VX450W. Cables are really embarassingly long, a problem I liked to have! The most impressive thing is silence. This PSU is almost unaudible, cpu fan is slowed down by fan management of the motherboard, now I can clearly ear the sound of the hard disk activity. Of course there's a price to pay, with the generic 450W PSU I could freely stealth fart, now I'm forced to a strict zero emissions policy... :(

Yeah I'd give the OCZ 500W a "best of breed" award. It stayed mostly square center in the benchmarks, sure, but the features are nice for only 80 bucks.

I really appreciate this comparison -- this is the mainstream segment that suits the needs of the large, non-hardcore majority. That may not represent the typical TR reader, but it certainly represents a lot of the people who ask me for component advice. Good to know there are some quality offerings in this price and wattage range.

I agree with the editor's choice picks, but I have to say I'd pay extra for the significant reduction in noise offered by the Silencer 750 -- 3 to 4 dB is a lot.

Of course when you're looking at efficiency the comparison is a little misleading, because given (say) a 250W load the 500W PSUs will be running at 50% and the 750W will be running closer to 25%. But the overall efficiency numbers for all but the crappy generic model are close enough that it really doesn't matter.