My comment on this review could sound like a personal attack which is NOT my intent. You need technical help to do this kind of review. Your faulted definition of power factor vs efficiency proves that you have NO basic knowledge of simple beginning electronics. Thus the rest of your review is completely suspect!

You should ask someone who is knowledgeable in electronics to review your stuff before you publish it..

Or maybe you could do a Google search for "power factor" to learn the basics of what that term means. Then do the same for "efficiency".

About the Neopower: I wonder if the modular design's extra connection (mechanical instead of solder) was introducing extra impedance and thus causing voltage drops across the board. After all, all three (four?) rails were a hair low, but had good ripple.

A simple rule of thumb on all support equipment be it psu or ups or anything else for anything else.... the more you spend on the thing its gona connect to the more you better spend on it relative of course to the general spendyness of what your dealing with.

Thus if your puter costs norm plus 200 then your psu prolly should be norm plus 25% in quality and likely price.

The trick of course being finding out what the heck is normal;/

Uhm, I have a little 'Elix' Powermeter and I thought, why not measure the voltages on my newly bought OCZ Powerstream. First problem; how do you measure? Just stick it in a molex? I know the basics of electronics, so 5v =red and black, 12v = yellow and black, 7v= read and yellow, but how do you measure the 3v rail? Is it even right to stick the ends of the meter in a molex or do I have to measure somewhere else?
So far I got 5.24 v out of the red and black and 12.36 out of the yellow and black, are these readings even close to reality, because thy seem so far compared to the article.

Great review, but there's one thing I'd like to see: how long will a system run Prime95 with each PSU before choking? I'd like to see just what the consequences of these numbers are.

Great review! A few minor points/nits:

- What exactly is the point of having adjustable voltages, anyway? Wouldn't it be better if the PSU makers just improved their QC to the point where all of the rails were spot-on when they leave the factory?

- Re your concerns about the fan grill on the Ultra X-Connect restricting airflow: How hard would it be to remove the existing grill and replace it with a standard wire grill?

- You might want to invest in a small digital postal scale, if you're going to make a habit of weighing components. ;-)

- I am a little bit puzzled by your concluding comments on the Silverstone. How can it be suitable for those who are "on a quest for the quietest system on the block", when it placed 3rd in the front-of-case noise test at idle, and dead last under load? I think your own noise level tests are evidence that radiating the heat from the PSU directly into the case and relying on case ventilation to remove that heat is not a winning proposition.

- My $.02 on the PFC debate: IIRC the wattmeters that the electric company uses to compute your bill are not/] very accurate when the PF is far from 1.0, and will tend to read low; this is one of the reasons the power companies get upset when a large customer has a PF drastically different from 1.0. If the wattmeter used for this article reads low when the PF is off, then it would be overstating the power usage of the PSU with the active PFC.

- Irrespective of how good the Silent Boost is, the actual idle CPU temp cannot be at ambient. Even the best air coolers have enough thermal resistance that there will be a temperature differential of at least a few degrees. The CPU sensor must be reading low. In this context it doesn't matter though, since what we're really interested in is the relative differences when using different PSUs.

One power supply line I'd love to see compared is PC Power & Cooling's units. I actually have one in my current system, but it's likely overkill. But I admit I don't know how MUCH overkill it is and whether it was necessary. [or maybe PP&C is actually nothing special but seems to imply it is, like Monster Cable or Prada].
Kudos to Damage for a great article! We just want more! :-)

Active PFC is important for any system that runs on a UPS.

As #46 hinted, "Looking at the passive PSU's 323W power consumption under load, it's clear that Silverstone's 300W rating is on the conservative side." is wrong. These PSUs are not 100% efficient.

I'd suggest that TR's staff take some time to read more PSU reviews by Mike Chin at http://www.silentpcreview.com and Lee Garbutt at http://www.systemcooling.com.

A great review as always. Excellent work on the graphs!

Shame there wasn’t a thermaltake in there though. Ohwell.

Here is MY experience. There are many crappy PSUs out there. Some better-built units use decent parts, but still try to be affordable. Antec and Enermax are primary examples of this second group. They are well made but still inexpensive. Then there are the units that are made well first, then made to cost second. An example of the third group would be PC Power and Cooling.

My experience is that Antec and Enermax PSUs will last forever if they are kept under reasonable load. If you run them for an extended period at closer to max load, they will fail. I have had it happen to multiple units from both companies. The difference between these and the cheapo units are that they are stable and non-spiky until the point where they really start to go. Moreover, when they go, you have plenty of warning because one or more voltage rails slowly drifts below spec over time. When it starts getting alarmingly low, time to replace the PSU. A cheapo might get spiky and give you inconsistent power, or it might just go in a flash.

The third group, typified by PC Power and Cooling, costs more than your average geek is willing to spend, but the component standards are even higher. These can generally "handle the heat" better than the second group.

Here is an example. I have a dual-processor Athlon server. It has an AMI MegaRAID controller with a bunch of SCSI drives plugged into it, a SCSI DAT library (6-tapes), and three IDE drives RAIDed. It has lots of fans to keep air flowing. This machine has already burned out two Antec TruPower 550 EPS (server) power supplies. I got tired of replacing them, so spent the extra bucks and went with PC Power & Cooling's ~500w EPS server PSU.

In the Antec case, the failure could be seen coming miles away. The 5v rail would continue to fall slowly over time. Eventually, it would get to the point where the machine became unstable, but lots of warning was given beforehand.

The PC P&C power supply has been absolutely rock-solid stable since it was installed. Every voltage rail is as precisely at the same voltage as the day it was new. There has been no measurable degradation, and the machine has been up for months at a time between reboots.

Such is my saga...

damn, I was wondering if I should even bother changing out my antec truepower with the PC power and cooling 510 watt I have sitting idle :)

I wish there would have been some tests with a system geared towards low noise. (Clearly, I'm not expecting to fit a dual Opteron system with a pile of 10k disks in a small, low-noise box, but still)

The noise measurements seem to indicate that the main source of noise in this case is not the PSU fan, but rather the CPU or case fan.

Even using the same setup, it would probably be possible to tweak the fan speeds (in other words, allow a higher case temp).

-fyo

I'd have loved to have seen the SeaSonic SuperSilencer 400AGX represented, also something from ThermalTake too in their Active PFC range too. I'm going to want to buy a new PSU for an Athlon64 server system I'm planning on building. So far the SeaSonic 400AGX is winning in my view ($90 from DTT computers too) since it's very quiet, very efficient, doesn't have a lot of jazzy crap that you're paying for.

Coolermaster is about to release a new power supply. Maybe you can suppliment this review with other manufacturers too.

Awsome review.
Those of us with a dual Opteron or any dual system thank you for using it to test.
95% of the time I get Antec, but I am looking at an Enermax now, so I am glad it did well.

Thanks again for the hard work.

I have to say I agree with #13 about confusion in the article over Power Factor and Efficiency:
Power Factor doesn't matter if you are a residential power customer, you get billed by how much power you use. Power Factor matters if you are a commercial customer, because you get billed by power used and power factor of your load, which is why you sometimes see banks of capacitors outside industrial plants to correct the power factor. They are cheaper than paying the penalty for the really bad PF.

I am concerned on the wattage readings - some meters read true power consumed, others are thrown by the power factor. What was this meter?

Efficiency is just power out / power in. You can get power in off a true power meter and then you need to measure power out to the board.

I also don't think the graphs are too helpful in just displaying the voltages. I would have liked to see how much the voltage differs as the load increases so you can easily see the difference between a supply that is just calibrated low but keeps the same voltage regardless of power and one that is calibrated better but sags when you start drawing power from it.

I also think the supplys should have been tested on a proper test rig and pushed to the limits, if they say they will deliver x amps on a rail or y watts on a combination, then test them to that limit and beyond.

A comment on the 12V ripple on the generic supply. Originally 12V was only used for drive motors and supplying the serial ports. If your computer only uses this then that supply is fine as the motors don't care about the ripple.. P4s introduced the concept of using the 12V supply for additional power for the CPU. In this case you are dependent on the motherboard for the quality of the CPU power conversion circuitry which may deal fine with that ripple, then again it may not and may cause problems with the CPU. Again it would be interesting to graph magnitude of ripple against load.

The weight of a PSU being a sign of quality comes from the idea of a quality PSU handling a lot of power and so needing big heatsinks. With differing fans this becomes less relevent.

Sorry for the length, but I'm a hobby electronics geek with a physics degree ;-)

Excellent review. There may have been a few points of confusion regarding power factor and efficiency - but hey, that is what the discussion is for - to bring these issues to light, and thus improve the testing in the future.

This is the best review of power supplies that I have ever read. I have only read one other of similar quality, but it is a few years old. It is good to see a thorough update, as modern high-end systems become power hogs.

Question: If the dual opteron (with several HD's) only draws 359 W (with the least efficient power supply) at the wall under full load, than what justification is there for a 550 W supply ? Large servers often have multiple power supplies to handle different components. I think the PSU manufacturers have brainwashed the masses into buying bigger (and more expensive) power supplies than are needed.

If the least efficient power supply draws 359 W at 75% efficiency, than that translates into 269 W. I thought power supplies are rated by their output capcity, not how much they are able to draw from the socket. If I am wrong, please let me know.

I actually just ordered an OCZ 520 a week ago, going to pick it up in an hour or so. That was based on another (equally good) psu article, don't remember where but the OCZ came out on top there.

Nice review TR, thank you. Great community response. Good looking out Jackson (#13).

Still like to see the $56 Aspire 520watt in there. Affordable, attractive. I wish I had an oscilliscope and the knowledge to use one.

I'm basing my next PSU buy on this (in fact I think I will replace mine). Had the TR a store... ;)

That was good stuff, hope to see more of this again, very helpful. VERY!

Nice work!

I guess I'll be considering an OCZ PSU alongside my usual Thermaltake fanboyisms when I build my Winchester box this winter.

Thanks for the heads up guys!

Good to see something other than the usual reviews around here. Thanks for listening!

Deleted - what rubbish I wrote :-)

Soon, digital tracking PS will make things more efficient and less power will be lost to heat. PS won't even have fans.

We'll, I'm also happy to see that the x-connect performed so well (even led in some areas) since I just recommended one to a friend!

Good review Diss!

Id imagine the reason the generic looks the way it does is its ment for a server where its likely needed to power 8 hdds some fans a gibit ethernet and loads of ram all of wich drain the 3.3 and 5 volt lines. Its prolly not made to run a high wattage cpu or gpu or specialy BOTH.

Thing to look out for some of those supposedly good performners in this test had rather weak 3.3 and 5 volt rails so if your into lots of hdds and fans and lighthing and blah blah blah... foom!

so what sort of damage can the massive variation in ripples cause to the system...if any?

Well its good to see the Ultra X-Connect didnt do to bad, im hoping to get one of those pretty psu's soon, Xoxide has window modded ones for like $119, and im taken by the modular idea, although, i might make my own atx connector....