Tom's Hardware Compares Power Supplies

Some guy wrote in to say "Tom's Hardware Guide takes a hard look at power supplies to find out if we are getting what we paid for. The results of the testing were very surprising." Very useful to anyone who has built their own machine from scratch or burned out a cheap power supply.

Any article that has a flaming PSU is good.

[Prince_Ali]

I hope a few of those people who pay $400 for the latest and greatest video card and $15 for a power supply read this.

Missing Test Equipment

[CMiYC]

I don't think these were fair tests. I would have liked to seen some oscilloscope measurements of what the voltages/waveforms looked like under full load. Being that we expect our DC power supply to delivery a DC voltage, even a novice can tell a "bad" output from a "good" output. Take two power supplies for example. A 300W and a 500W supply. (For numbers sake, let's say they only deliver 5V to the load. No +12v, -12v, etc). If I max load the 300W supply and it is delivering a clean 5volts, that's a great supply. But if the 500W supply is spiking or has considerable noise with a 300W load, who cares if it runs up to 500W?

To me that's almost more important than if the supply shuts itself down or not. Which, by the way, is still a nice FEATURE.

You get what you pay for

[Pedrito]

This applies particularly to power supplies. Sure, CPUs and memory, but the prices aren't nearly as fixed as they are for power supplies. Really, with power supplies, the price range doesn't vary much and the good ones tend to cost (though there are some decent ones for decent prices).

Before I came to my company, they bought a bunch of no-name PCs. There must have been a motherboard flaw that caused them to burn out power supplies and they kept replacing them with cheap supplies which couldn't handle whatever the motherboard was doing, and they would burn out too. Out of about 8 machines, I think we went through 14 power supplies in two years.

You'd always hear, "What's that burning smell?" "Did you check the back of your machine? I think that's smoke from your power supply."

I ALWAYS get a decent supply and have NEVER had problems, even when I lived in Mexico and had pretty questionable electricity.

Tomshardware

[Anonymous Coward]

A single floppy connector, as supplied by Engelking and Coba, is not enough. You will need at least two of these.

Haven't given them a thought in over a year. All their "reviews" tend to shine highly on the products that fit "their" personal views. I mean jesus, any place that thinks you should have 2 (two) floppy drive power connectors is a little behind the times. Most people don't even use 1 floppy, let alone 2. And for all you people who weren't aware, a reviewer is supposed to enter a situation unbiased. Tom's hasn't started a review unbiased in easily 2 years.

No PC Power and Cooling?

[LordNimon]

The only decent third-party manufacturer of power supplies that I (and I presume most people) have every heard of is PC Power and Cooling [pcpowercooling.com]. It doesn't appear that this article covers any of their products. Am I missing something?

Only one of each tested

[f97tosc]

That is not a very large sample. I wonder how much specs vary between individual units of the same make and model.

In particular, it would not surprise me if there are unit-to-unit variations in noise and the power at which they give up.

Tor

How good is Enlight?

[antdude]

How is Enlight power supply unit (340 watts) in a gaming and workstation machine? I know they're a bit loud, but how's the performance and reliability?

PS Diagonistics?

[Masem]

I'm surprised that in this day and age, where your power supply is becoming one of the more critical components to keep track of (along with the CPU and GPU temperatures), that there's yet to be a way to monitor the performance of the PS with hardware/software monitors, short of wiring your own. That is, just like you can monitor temperatures and fan speeds with most modern mobos, the power supply is completely independant of this. Yes, it would require some standardization of how that info is sent and a plug on the mobo (most likely situated near where most USB/KB/M cutouts are as to avoid a 20ft wire to get it to the northbridge site), but it would seem to me to be really useful information to determine the PS health beyond the current hope-n-pray methods...

Re:No PC Power and Cooling?

[mackstann]

you are not missing anything, unfortunately.

this review is pretty lame IMO, they failed to review: both sparkle and pc power & cooling, probably two of the highest regarded names in power supplies, while at the same time they reviewed such no-name junk as "engelking", "coba", "channel well", "scs", "task", and "levicom".

and the summary is also quite worthless. basically they recommend 3 no-name brands as the "winners", no one is going to buy these due to availability and/or fear of no-name brands, while THG also recommends antec psu's to those who cannot find the others, which is also stupid because people already buy antecs in droves.

also we have the issue of rebranding. many power supplies on the market are simply rebranded no-name power supplies, how do we know that just because one "herolchi" power supply tested well, all others will also be top quality? the different herolchi models could be from completely different manufacturers!

Could have done better...

[KC7GR]

For my part, I would have liked to have seen THG use a true electronic load for their testing. Something from Transistor Devices [tdi.com] 'Dynaload' line would, I think, have been a much more accurate (if more expensive) choice than a box full of power resistors.

They should also have used a good O-scope to take a look at the power output waveforms while under load. I've seen a number of cases where a switching supply will look perfectly clean under low-to-medium loads, and then start to spike and freak out under higher loads. 'Tis a nasty thing to behold, and it can cause problems that can drive techs who don't know what to look for absolutely batty.

Also, others have mentioned that PC Power and Cooling was left out of the review for reasons unknown. I would guess that it was price. If so, all I can say is "How highly do you value your hardware?"

Clean and adequate power is the ONE factor that can cause more woes than any other. You can have the slickest quad-processor-super-Linux-cluster-RAID-whatever on the planet, and it won't do you one whit of good if you've got dirty power feeding it.

Re:Non-custom built power supplies poor?

[Tailhook]

It's a question of usage patterns and necessary capacity. The common Best Buy PC used in a typical manner does not tax a garden variety Chinese PSU enough to ruin it.

A SMP box with a Gig or more of RAM, used to compile kernels, run FPSs at high resolution, host a couple extra drives of various sorts, get frequently booted between multiple OSes (startup loads are extraordinary,) run benchmarks, and basically do a bunch of other crap, will need a LOT more power. I fried an Antec 300W PSU in 3 months like this. Give yourself a fright and watch the case temperature during a FreeBSD "makeworld" sometime.

Tom's caters to people that push high end hardware to it's limit. You're basically reading hotrod magazine and wondering what's wrong with your Accord.

Re:Fireworks

[Oloryn]

One leaves the computer plugged in for the ground. Unplugged, you won't get electrocuted, but the box won't be grounded either. You have to pick one, or find another way of grounding the box (which realy isn't that hard)

For that matter, if you're concerned about static zapping computer components, it's typically sufficient to make sure you are at the same potential as the computer by using a wrist strap connected to the chassis.

Re:sparkle power supplies

[XorNand]

I second the recommendation for PC Power and Cooling [pcpowercooling.com]. I got tired of going through a PS every 6 months, having themburn-up or worse; the bushings would start to fail and it would sound like a turbine engine until it warmed up. As far a price goes; it's a matter of perspective. I think I paid around $200 for my high-perf 350W supply about 5 years ago. To date, it's the only piece that I haven't upgraded or replaced in my system (including the case). Spend twice that amount for the latest bleeding-edge graphics card and see if you still have it half for half as long.

How About Using LESS POWER?

[ink]

Am I the only person who is sick of computers requiring such obscene amounts of power? Newer machines have fscking radiators on them for $DIETY's sake; what's next? A heat-pump that sits outside my house to keep the environment nice and warm? In 1995, 250W was a nice, big power supply. Then, 300W, and now 500W comes along -- other consumer electronics are becomming more efficient (monitors, televisions, refridgerators, air conditioners, etc.), but computers just keep wasting more and more power.

Re:Missing Test Equipment

[mark_anderson]

CMiYC has an important point. There's a lot more to a power supply than simply providing lots of DC current. Tom's Hardware used a constant load. Computers, especially CPUs do not form a constant load. A cpu may increase its consumption by a factor of 10 almost instantaneously. This can happen everytime the scheduler goes from the idle loop to running a cpu intensive task. The motherboard regulation will absorb some of this, but not all. The PSU must be able to respond to these surges without significant ripple or spikes. This requires good capacitors, and may require tuning the switcher frequency to improve the response.

Re:don't you think?

[blincoln]

if you like pseudotechnical hogwash, incestuous relationships between advertisers and content providers, and meaningless benchmark data.

While I wouldn't flame them quite so hard, I have to agree with the spirit. THG lost all of my respect with their handling of that "hot contraband" P4 article. If I can't trust a website to tell me when they're Photoshopping their BIOS screens and shots of CPUs, how can I believe their benchmarks are genuine, or their reviews unbiased?

Re:Marketing lessons from audio amps?

[Lumpy]

and it's ratings were accurate.

It was more than likely rated at 500watts Peak and did not mention a THD rating at that wattage.

Car audio is usually marketed at truthful levels. Pyramid (I think that's how they spelled it) amps were usually rated at thousands of watts. but if you look it is rated as peak not RMS and it's THD was at 5%.

now an amp that sounds great is rated at 0.05%THD at RMS watts. a 100watt RMS at 0.05%THD will knock your socks off with a pair of 15" woofers in an isobarik enclosure.

the moral? read and UNDERSTAND the specs... car audio makers bank on the fact that the large number of cunsumers are too dim to understand the markings on the box.

Re:Fireworks

[Anonymous Coward]

corollary: never let a hw guy write software if you ever want to have a chance at understanding the code.

corollary2: all hw guys claim to know C.

Re:How About Using LESS POWER?

[_|()|\|]

Am I the only person who is sick of computers requiring such obscene amounts of power?

You have a choice in the matter. If you want top of the line performance, get an Athlon or a Pentium 4 (50 - 70 W) with power-hungry components. Otherwise, you can get a Celeron (25 W), a K6-III+ (20 W), or a C3 (10 W) with one hard drive and a video card that doesn't require active cooling.

I bought a little FlexATX bare-bones system that's a little smaller than a bread box. I put in a 5,400 RPM drive and a Celeron 850. The only fan is in the 100 W power supply.

Mutton dressed up as Lamb... as per usual...

[Shanep]

There is no mention in this article of ripple performance for these PSU's under their rated loads.

Ripple is the amount of AC left in the DC output of a DC power supply. Real engineers and technicians measure this and take it into consideration to assure the reliable and healthy operation of the equipment being powered. There is no point in having a PSU that can deliver the required amount of power if it is also delivering the parts in the computer noisy power that may lead to instabilities.

The PSU's that actually were able to deliver more than their rated power, may have in fact been designed so that they actually deliver low ripple power at their rated levels. With power beyond that starting to show what the designers would deem, unsatisfactory ripple levels (Ripple becomes more apparent with higher loads).

A quiet (electrical) supply is a good thing for computers of any size and seeing an article at Tom's omit this amongst pages and pages of a "test" comparison does not surprise me.

Blah blah blah. People who know better, don't read Tom's, they "do it" themselves, properly. But the chance to test 21 different PSU's is something few geeks can do, so Tom ought to get things done correctly if he is to pass his site off as a valuable technical hardware resource.

But what I think is the real killer, is that Tom tests the noise levels of these PSU's, but not the electrical noise, the audible! Which kinda shows in a glaring manner the level of technical prowess his site staff and readership posses. Hell, they had multimeters, how hard was it to at least set them to AC and read the amplitude of the ripple!

"Test results in detail" my arse.

I'm not being picky BTW, ripple testing is a must do in PSU design and testing for most applications of a DC supply. Proper "test results in detail" would have included oscilloscope printouts of the ripple, IMHO.