Slashdot Mirror
Five Power Supplies Compared
EconolineCrush writes "Tech Report has done up a comparison of five high-end power supplies that looks at actual voltage levels and AC ripple content. The article also takes a look at environmental factors like noise levels and each power supply's impact on system temperatures. Think power supplies with like wattages are created equal? Think again."
Is there something that these power supplies contribute towards overall system stability that "cheap" ones don't? Are they really worth the money?
The article was very good at measuring everything measurable about the power supplies, but didn't answer the question "Why would I want one of these?". So why would I?
My other car is first.
I seriously tire of "tech" reviews on stuff like power supplies, roll out keyboard drawers, cd holders, etc... This is the third "tech" article about power supplies I've seen here in a year.
Here's all you need to know:
Pick up two of the same rating, different brands. The heavier one is better - more windings on the coils and better components.
The end.
I don't need no instructions to know how to rock!!!!
http://www.tomshardware.com/howto/20021021/index.h tml
:)
21 power supply tested here
Before I read the article my guess was that Antec would win...and I was right. From the low end to the high end these guys have got their stuff together.
Most of the cases we buy come with Enlight power supplies (they are Enlight cases after all for the most part). Although these Enlight PSes seem to be ok, I always replace them with a nice quiet reliable Antec when they are going home to me or to my family. I also recommend putting an Antec PS in to customers who buy the biggest, baddest gaming PCs.
The simple fact of the matter is though, that most folks don't really need a 550 watt PS. A 350 watt PS will more than handle the load of most average consumer PCs. I do dread opening up an e-Machine or various other "value" (aka cheap ass P.O.S) PC and seeing a 130 watt PS running a P4 CPU. *shudder*
Duris MUD - The best pkill MUD. Ever.
While all of these look good and they all have some pretty spiffy specs, it would have been nice to have seen some reliability test scores in there. A flaky power supply is a hard thing to track sometimes, and knowing which ones have the best chance of running reliably for the longest time period would be handy information, especially for those of us who have fought with bad power supplies at one point or another in the past.
Be excellent to each other. And... PARTY ON, DUDES!
Good PSUs contribute a lot towards system stability. For example, a moving head in the HDD causes current transients (so does a CPU switching between normal operation and power saving mode). Bad PSUs have huge voltage drops during these transients, good PSUs can buffer them quite well. These transients can cause anything between nothing and total system crash.
Also, the ability to filter noise out of the AC helps with stability...
"Aesthetically, there's not much to see with the SilentX; it looks like, well, a power supply."
I've had so many powersupplies dying on me that it's not even funny. What I want is a PSU that delivers the promised effect, for at least three years. That would be the day.
Might be hard to benchmark that, so anybody got any tips for brands?
How small a thought it takes to fill a whole life
AnandTech also just had a PSU roundup here. The watt numbers and some of that aren't the best from AnandTech's review (read some criticism of it here), but overal it's a good roundup, especially comparing heat and noise.
As the article concludes, Antec is the best option available. I run a server with a 300 watt Antec powering the system, and a separate Antec 400 watt running ten of the eleven hard drives. The voltages stay tight and the supplies stay cool. The 500+ watt models are expensive, but the $69 I paid for the 400 watt is well worth it when you are protecting 1.1TB worth of drives.
:(
On the cheap/lower power side, I've had great success with Sparkle and Enlight (250-350 watt) supplies. Priced in the $22-$40 range, these are great for "normal" systems.
I definitely recommend you stay away from the cheap stuff that comes in $30 cases, though... you'll see why when the supply dies, or shorts taking the mobo with it.
Read this page from AnandTech's PSU roundup. The only thing that was changed between tests was PSUs, yet over a 6 hour time there was a range of only 1 up to 7 memory errors. Just one possible indication of how clean the supply is. Also there's the added benefit that it probably won't die in a month or two (happened to me, not too pleased when it did)
Where can I find this cheap gas now?
Slashdotter are stupid and biased.
anandtech reviewed 18 units on july 31.
Interesting there is the memory test: they show that stable power gives less memory errors with memtest.
And here is clickable tomshardware:
Toms burns up some power supplies
In a cheap power supply, you can get inconsistent voltages, crazy transients, crosstalk, and if the power demand from one line goes up it can drop the wattage/voltage on the other lines. Cheap power supplies are also frequently noisier (sound too) than high end ones and run less efficiently (read: hot) than better designed, more expensive power supplies. Think about it: your computer operates because of well controlled voltages. If your voltage drops by 2V, some transistors will go into their linear range and cause crazy crazy crap to happen.
See my reply here.
The reason they check the voltages so closely is that one you start falling out of the 5 or maybe 10% tolerance zone for many components, over-voltage will cause overheating, lockups and early failure, and undervoltage also frequently causes lockups and occaisionally failure.
Also, some supplies give you a total wattage without breaking down where those watts can go. When you're dealing with processors that pull 80 watts at peak, you REALLY don't want a cheap supply that is busy sending all available watts to 5 and 12 volt channels to power drives.
Another thing to consider when buying a case.. the PS they put in cases are the CHEAP kind, unless they specify what kind it is, I generally expect to replace it within a year. A few years ago I had one of the dual socket370 BP-6 boards, it refused to boot on the PS I had that came with my case (Enlight none the less). I swapped it to a sparkle 300watt and have had no probs since.
This was also recently covered by Tom's Hardware, and earlier by a few other sites. The sparkle and HEC normally blow away the rest, with their 250w beating the specs for most 300+w, and even being able to hold 300w operation themselves.
just my $.02
Tm
Support TBI Research: http://www.raisinhope.org
If you've got weak voltages on the PSU rails, it can kill your HDDs. Some people lose drive after drive and never consider that their voltages are out of spec. Also, if your cheap PSU shorts on the DC side, say goodbye to your drives and maybe your motherboard and everything plugged into it.
You've got to be trolling. On the off chance that you aren't:
/. intro noted case noise and heat output into the case.
The article started right off by saying that system stability can be affect, that the stability of voltage levels and the amount of electrical noise varies greatly. It also noted that the power supplies distribute their power differently among the various output voltages The
The effects on the CPU, chipset and RAM of electrical noise and/or 'brownouts' of voltage dropping below specs should be obvious. I've seen several systems go instable because the 5vsb line, or some voltage branch like the USB line couldn't drive the attached components. What good is having 200 extra watts you don't need at one voltage, if the PS goes flaky at full output and real usage on another? A lot of power supplies that do fine (or almost fine) on a bench or at 50% of their rated current draw in the real world will flake out occassionally at 85%. A few milliseconds of flaking out ever several hours can turn a dream machine into a nightmare.
Hook an oscilloscope to distal power traces on the motherboard (not near the power supply, and depending on your supply, you can see some pretty ugly stuff as peripherals/cards switch on/off. Sure, a good motherboard should have plenty of well placed filter caps, but on a fully loaded system, you can *see* how adequate they sometimes aren't, if the power supply doesn't supply great power in he first place. It's possible to design very rugged and tolerant motherboards (e.g. military), but in the consumer market, price competition is so tight that boards are often revised in mid-production to save one or two caps.
I'm not saying top-of-the line is always best, but bottom of the line is pretty much asking for trouble down the line. Most people 'add and test' when they build (or expand a system with use), but the culprit may not be the card you just added; it could be the power supply you 'vetted' up front.
Why is it that every time these "high-end" power supplies get compared, the most high-end one always gets ignored? PC Power and Cooling has long manufactured the world's best power supplies. They're the Ferrari or the Moto Guzzi of the power supply world. Yeah, lots of Asian manufacturing firms make OK power supplies, but PCP&C's stuff is the only company that makes boards that the major motherboard manufacturers highly recommend and use exclusively in their own tests. Why does such an obvious high-quality product always get ignored?
Rail wattage is also important because power supplies have 12, 5, and 3.3 volt supply rails. If the 5 and 3.3 are sharing the same rail then you are limited to the wattage from that single rail. Antec was the only power supply to have separate 5 and 3.3 rails. If you don't have enough power going to your components under a heavy load, your system will crash. Think of the power supply like a carburetor. If it's running too lean or you put one 2 barrel carb on a Hemi, it won't run right. If your power supply can't get enough voltage or current out to your components, or you put one on that's too small, your computer won't run right.
Needless to say, I'd buy the Antec. Separate rails for 5 and 3.3 plus the second-lowest ripple make it a good choice for stable power.
From the article:
Would you dare put cheap gas in a Porsche?
Am I the only one who answered "Absolutely; gas is gas"?
"Goodness me, how unlike the FBI to abuse the trust of the American public." -- The Onion
Why PSU reviews? As you so effectively illustrated, there are a lot of people out there who think that one power supply is as good as another. There are people who complain about kernel panics and blue screens and swear up and down their hardware is fine, but still have the same $3 power supply that came with the case.
Ask a hardware engineer about the importance of clean, consistent power. I'll bet you dollars to donuts he won't say "buy the cheapest supply you can find, 'cause big PSU's are only good for powering LAN party foolz' casemods."
If a job's not worth doing, it's not worth doing right.
Who told you that you should be running your power supplies at close to their maximum load? The guy who was trying to sell you the 200W power supply for a P4? Running at close to the maximum load constnatly will reduce the life of your power supply and make it run hotter.
Also, your surge supressor won't stop normal power spikes, they are only designed to stop grossly over voltage spike like you get with lightning strikes. The worst part is that many of them don't even do that. Did you know that most surge supressors, once they're "blown" will fail into an "on" state. It's amost impossible to check to see if your surge supressor is still working.
I read the internet for the articles.
Well, I have found PSU problems. And the ulgy thing about it is that the PSU is the last thing I would have thought...
A couple of years ago I put together a decent system, TB 1GHz, 256MB DDR, ECS k7s5a mobo, etc. And a crappy tower/psu cuz I spent too much so I cut some corners on case/keyboard/mouse/etc. BIG mistake.
I started having problems with the computer. Random crashes and all. Damn, I think. Somebody suggests the memory is faulty. So I roll out memtest86, and sure enough, both memory modules are faulty! On a hunch, I try them on another computer, and they pass ok. So, the CPU must be fucked up. On another system, the CPU passes the test. So it must be the frigging mainboard. Someone suggests I try a different PSU, since the Athlons are known for their watt gluttony. Put in a new PSU, a bigger and decent-er one, and all problems evaporate. No more memtest fails, no more freezing, nada.
I'm not a power supply designer, but I do have some experience with system power supplies and their affect on system operation/reliability. If I were deciding upon a power supply for my system (or product), I would carry out the same testing as in the article, but also measure the following four parameters:
1. Initial Power Up overshoot/ringing/stabilization. I would hope the supllies powered up with a basic RC curve "POWERGOOD" becoming active when each of the supplies are within 1% of their targets.
2. Transient response. This is different from the "Load" test, it would look at how the power supply worked when it went from minimum load to maximum load and back again. Say starting up the disk drives, CD-ROM and change the fan speed at the same time.
3. Transient response across supplies. What happens if there is a large transient on another supply. The different power outputs in modern power supplies are not as separate as you might think.
4. Power down characteristics. Again, this should be a smooth RC curve with no overshoots or ringing. The high power positive voltage outputs should never go negative.
The first and last parameters will be an indicator of how "gentle" the power supply is on the components and whether or not there is any danger of having them overstressed. The middle two parameters would indicate how reliable operation of the PC would be and whether or not you would get power supply induced lock ups or glitches.
Power supply design is more art and component management than strong engineering application. Modern PC power supplies really are a result of iterative cost reduction and learned experience. A lot of "common sense" ideas are just plain wrong when applied to high current output switching AC/DC converters: I have learned that heavier is not always better and is often an indicator of an inefficient design. Fires are not uncommon in PC power supply testing and development and choosing the best power supply design is often a case of figuring out which company could best understand what the ashes were teling them.
myke
Mimetics Inc. Twitter
Er, not necessarily. Since heat rises, and the power supply is at the top, it shouldn't heat the case much anyway. Having recently switched from a loud PSU to a Zalmon power supply and lowered the voltage on my case fan from 12v to 7v (to make it less noisy), I've noticed only a minimal increase in temperature (still waybelow any max limits) and a blessed release from buzzing noises... I can now sleep with my PC at the end of the bed switched on, and more importantly, so can my girlfriend (admittedly not the usual review criteria for Slashdot ....)
P.
http://www.anandtech.com/showdoc.html?i=1841
Does the name Pavlov ring a bell?
I'm an MIS guy for a small company (10 people, 20 PCs -- go figure...) and I always look at PC power and cooling supplies as well as other brands when I'm building machines. I think they make great server supplies or swap in replacements for older machines at the very least.
I have also used those guys for obscure CPU cooling fan options (try to find a quality replacement CPU fan for a Pentium Pro 200)! They stopped stocking them, but offered to make one up for a very reasonable cost -- I went with a different solution, but they were quite helpful. I have purchased several CPU fans from them and none have yet died.
I usually go with Antec power supllies for new workstations because, in addition to running well, they come standard in good Antec cases that I'd want for a workstation anyway.
I'm thinking about it, therefore I might be.
A/C ripple would be 60Hz (or perhaps some harmonic.) A 10ms sample is _woefully_ too short to see it.
... it is my understanding that switching powersupplies shift the frequency to something substantially higher (1-10k Hz region.)
... approximate amplitude was the only conclusion you could draw, but to say "look, A/C" ripple noise is just plain silly ... especially as the A/C ripple would be so much larger than the 10ms sample duration.
Also, of more interest in a switching power supply than A/C ripple would have been the ripple from the power supplies own oscillator
Take a long (say 1 second) sample at a decent resolution (at least 120Hz for a A/C ripple, perhaps 100khz if you care for switching ripple) perform an fft, and look for the spikes at 60Hz and whatever the powersupply switches at.
What they were showing was meaningless noise
That's true. The hdd motor and head positioning is usually running on the 12V supply and its current profile is on/off at fairly high frequency.
We used to estimate power supply quality by weight. The heavier, the better, since it meant they had more iron. Bigger transformers = better magnetic storage = better voltage stability. Now the switching frequencies are high enough that you don't need big iron cores. But you do need a switching frequency that is a lot higher than the load current frequency. Otherwise the 12V won't be stable. Not that they will tell you the switching frequency in a spec.
300 watts is more than enough for you.
Wattage isnt as important as the individual amperages that can be delivered on each line, though.
You want 2A or so on the +5VSB line, for example, if you want to use wake-on-lan or wake-on-keyboard.
You want a strong 12V if you have a ton of neon lights and bullshit like that. P4s and (i think) Athlons use the 12V line as well.
Drives used to use 12V to power their motors, and 5V for logic, but they all pretty much use +5 these days.
You can look at the drives to see what they need.
It takes all of maybe a half an hour to see what the amperage requirements for each component is, and then find an appropriate supply.
You see little cube PCs with Radeon 9700s and P4s running with 200 watt supplies, and its perfectly fine so long as the 200 watts is going to the right places. PSUs also run more efficiently at full load too, so it's cheaper on the electric bill in the long run.
I don't need no instructions to know how to rock!!!!
Which one will give me the best framerate in Quake?
Outdoor digital photography, mostly in New Engl
I don't know if you'd want one of these top-of-the-line power supplies, but you definitely want a decent one.
About three years ago I bought a case without checking the power supply in it and after about a month of operation my mainboard died. I blamed the mobo (it was also a cheap brand), and replaced it with a really nice one. That one lasted about a year, but was really flaky the whole time, especially the onboard Promise ATA100 IDE controller which had so many errors that I stopped using it. When I decided to buy a new machine, I bought better stuff but I (foolishly) replaced the mobo in that system yet again and gave the thing to my wife. Where I'd seen minor instability and annoying failures under Linux, she saw daily bluescreens with Win2K. Finally I bought her a new power supply and all of the problems went away (well, she's still running Windows, but I'm working on that ;-) ).
So, at the end of it all, I'd say the $20 I saved on that cheap power supply cost me two motherboards.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
I just scanned the article, but I saw no mention of testing for line regulation. Maybe I'm just old school, but that used to be an important factor. Oh yeah, just thought of another one. Home users might be interested in knowing just how much noise that power supply is injecting back into your mains voltage. Switching power supplies are noisy little beasts.
Indeed, I was disappointed that their testing regime didn't include any disk seek stress tests; a test which forced two disks (or more) to simultaneously seek from track 0 to track N would would exercise the PSUs' transient capacity really well.
Many years ago, a development system I was using had a cabinet with four disks in it. Every once in a while, during parallel makes, all four disks would spin down simultaneously. Eventually, we discovered that if all four drives were told to seek simultaneously (easy to do on a SCSI bus), the resultant load on the 12V line would pull it out of spec, the power supply would shut down, and the disks would spin down (releasing the overload and allowing the power supply to come back up, hiding the evidence). Since this box was a kludge, we "solved" it with a big, fat capacitor on the 12V line (next to the drives) to handle transients. (Which probably reduced the power supply's lifetime due to power up transients, but who powers down development systems?)
Modern disks do draw less transient current during seeks, so this isn't quite the issue it used to be, but it is still a source of stress they ought to have checked.
Well lets see, I had one of those cheap no name powersupplies in my last system. One night it blew up. That in itself was pretty cool, had fire and shit shooting out the back of the case, smoke coming out all the holes. Now for the uncool part, every thing in the system was toasted. About 1500 bucks worth of system up in smoke.
I guess it could happen with any powersupply but in this Beast I bought the one they recommended last year.
Supporting World Peace Through Nuclear Pacification