PSU Guide

[Swamp Monster]

Now back to the real world:
1) instead of working or gaming do we have to watch a multimeter, just in case? no..please!!!!!
2) so I take the hassle to do the multimeter part, let's say, once: what benefit do I get?
3) can we avoid funky voltages? not really
4) can we enjoy a 'smart' monitoring? apparently it's all lies, lies, lies, so why worry? just keep your backups up to date...

1) If you want to see if similar voltage spike exists in real life, then you need to watch it closely for a long time to see if it happens or not.
Note that you might not see the spike with digital multimeter even if it exists, if it happens fast. With analog multimeter chances are better, but you really need to watch all the time, because you can easily miss it if you look away for a second. Better to use oscilloscope here, but that is in different price category.

2)If you connect a multimeter once and do a few tests - gaming, other stress tests, you will see how voltages changes with load. Then you will know how good or bad voltage readings are and if you need to worry about them or not.

3) if funky voltages exists, then you can take it back to the store.

4) It's not that simple. Monitoring depends on programs too. For example HWmonitor reads 10V on 12V rail, but Aida64 reads 12.096V. Generally if readings are wrong, then they are really wrong. If you find a program that reads monitoring chip's data right, then I don't see problem here. Of course it can be few mV off the real life, but it doesn't make it completely useless. <This is only my opinion, but on my PC I don't see monitoring chip to be completely wrong, only programs that reads data from it.

 

[westom]

Now back to the real world:
1) instead of working or gaming do we have to watch a multimeter, just in case? no..please!!!!!
2) so I take the hassle to do the multimeter part, let's say, once: what benefit do I get?
3) can we avoid funky voltages? not really

Once a monitor is calibrated, then the monitor can alarm for voltages that still keep the computer working but are unacceptable. That are typically a precursor to future failures. And short occur rarely.

A monitor typically does not report useful information voltages until voltages are first defined by a three digit multimeter. Only the few who actually do this stuff (ie posted datasheets) would understand that.

Most voltage problems that cause crashes, maybe once a day or less frequently, are actually happening constantly. Problem is constant. Resulting crashes are intermittent. Therefore a multimeter will identify the problem at any time.

Again, a computer can still boot and work fine when voltages are defective. A defective power supply can still boot and run a computer. Crashes may be intermittent. Reason for those intermittent crashes exist constantly. Only a multimeter would identify that problem before it causes a future crash or other failure.

You don't monitor anything constantly. Take a meter reading today to see why intermittent crashes were happening yesterday. Funky voltages are that easily identified and eliminated. You are not having a spike once every three days. You would be having a problem, constantly, that causes a crash maybe once every three days.

A monitor simply make possible discovering a funky voltage sooner. Swamp Monster better defines reality and how voltages work (ie under changing load).

Better techs install a new supply, maximize the computer load, then measure those supply voltages withy a multimeter. How to identify a defective new supply long before it causes computer crashes or the warranty expires.

 

[streetfighter 2]

Only a multimeter would identify that problem before it causes a future crash or other failure.

An oscilloscope would be far more effective. Defective/cheap power supplies often have higher ripple, which won't be easily detected with a multimeter.

During load it is normal for the 12V rail to drop a small amount and ripple increases. Higher ripple under larger loads will not be easily observable with a multimeter but can still result in system instability and/or component damage.

 

[cdawall]

Once a monitor is calibrated, then the monitor can alarm for voltages that still keep the computer working but are unacceptable. That are typically a precursor to future failures. And short occur rarely.

A monitor typically does not report useful information voltages until voltages are first defined by a three digit multimeter. Only the few who actually do this stuff (ie posted datasheets) would understand that.

Most voltage problems that cause crashes, maybe once a day or less frequently, are actually happening constantly. Problem is constant. Resulting crashes are intermittent. Therefore a multimeter will identify the problem at any time.

Again, a computer can still boot and work fine when voltages are defective. A defective power supply can still boot and run a computer. Crashes may be intermittent. Reason for those intermittent crashes exist constantly. Only a multimeter would identify that problem before it causes a future crash or other failure.

You don't monitor anything constantly. Take a meter reading today to see why intermittent crashes were happening yesterday. Funky voltages are that easily identified and eliminated. You are not having a spike once every three days. You would be having a problem, constantly, that causes a crash maybe once every three days.

A monitor simply make possible discovering a funky voltage sooner. Swamp Monster better defines reality and how voltages work (ie under changing load).

Better techs install a new supply, maximize the computer load, then measure those supply voltages withy a multimeter. How to identify a defective new supply long before it causes computer crashes or the warranty expires.

why do you continue to argue this i am going to post this at point blank and if you cannot understand it i would prefer you PM me instead of ruining a perfectly good PSU guide thread with nonsense.

There is no way to properly tune ANY windows, linux or mac based program to PROPERLY read ANY chip on ANY motherboard. YOU do not even properly understand how an ADC works in a nutshell input is an ANALOG voltage and output is a DIGITAL "1" or "0" ie a series of highs and lows. not some rectified voltage like you posted earlier. On top of that there is NO multimeter that can actually read constant dips and spikes i don't card if it s a 50 decimal. The only proper way to test ANY powersupply would be using an Oscilloscope along with a load tester. There is no computer hardware there is no mystically "fixing" some program there is a physical hardware based oscilloscope and something to properly load a powersupply.

THE END NO MORE ARGUEMENT!

 

[westom]

An oscilloscope would be far more effective. ... Higher ripple under larger loads will not be easily observable with a multimeter but can still result in system instability and/or component damage.

Yes, when designing power supplies, then test equipment is $thousands. And most every layman could not use it. But any layman, even a junior high science student, can have more than sufficient information using a multimeter. Excessive (a word different from higher) ripple voltage is obvious with a multimeter due to how a multimeter works. Details well beyond this scope. An OP takes numbers with a multimeter. Then posts based in wild speculation need not appear. Replies from the few who actually know this stuff come with firm (without doubt) recommendations.

A reply is only as useful as facts first provided. No numbers from a multimeter means those who actually know this stuff are silenced. A post with multimeter numbers means answers that are definitive - without doubt. Many important facts (not just ripple voltage) can be identified with one minute of labor and that meter.

 

[streetfighter 2]

Excessive (a word different from higher) ripple voltage is obvious with a multimeter due to how a multimeter works.

Unfortunately this isn't true. The rectifier output of your typical PSU is switching at >20kHz. Only a very nice multimeter (which costs as much as a decent oscilloscope) is going to reliably catch "excessive" ripple, where I'm defining excessive ripple as >±0.5V.

test equipment is $thousands. And most every layman could not use it.

There are quite a few ways to acquire an oscilloscope for less than $400. The best are bought secondhand from colleges/universities. More commonly people buy USB based oscopes that start at around $200. If you can use a multimeter you're overqualified for using a typical oscilloscope , and quickstart guides are readily available.

A post with multimeter numbers means answers that are definitive - without doubt.

Unfortunately this isn't always true. A standard multimeter just isn't going to reliably detect ripple.

For anyones reference:
http://www.hardwaresecrets.com/article/Why-99-Percent-of-Power-Supply-Reviews-Are-Wrong/410/1

 

[westom]

Unfortunately this isn't true.

Your numbers are correct. But excessive ripple voltage (ie > 200 mV) is detected by a multimeter because of how a $17 multimeter works. Which is why some failures were so quickly identified with only a meter and a maximum load - due to how a meter works and how supplies regulate.

Yes, an oscilloscope is the preferred tool for those who have $thousands in test equipment right there next to a computer. Layman must use something that only identifies most power problems and takes a minute of labor. Resulting numbers means a layman is assisted by those who actually know how power supplies and motherboard monitors work.

erixx - it is a power 'system'. Supply is only one component of that system. Multimeter numbers report on everything relevant in that system - not just on a supply. Whereas the motherboard monitor can identify typical problems before a failure results. A multimeter can report numbers that define everything in that ‘system’. What engineers were doing even long before the IBM PC existed.

Calibrating the motherboard monitor simply creates an alarm that can identify problems when the computer is still working. How to eliminate some failures when most convenient for you.

 

[streetfighter 2]

Your numbers are correct. But excessive ripple voltage (ie > 200 mV) is detected by a multimeter because of how a $17 multimeter works. Which is why some failures were so quickly identified with only a meter and a maximum load - due to how a meter works and how supplies regulate.

I own several multimeters and I have actually tried to measure ripple with them. I have 2 Flukes, 1 Extech, 1 Radioshack, 1 Harbor Freight POS and one automotive analog multimeter. None of them were able to confirm the 0.6Vp-p ripple (measured with a $100 oscope bought off craigslist) on my LPG2-43 (630W) power supply with a 250W load of 12V lights. (The LPG2-43 can only supply 26A on the 12V rail so 250W was a significant load.)

Perhaps you can tell us what $17 multimeter you are using to get your results.

Yes, an oscilloscope is the preferred tool for those who have $thousands in test equipment right there next to a computer. Layman must use something that only identifies most power problems and takes a minute of labor.

http://www.easysync-ltd.com/product/520/ds1m12.html
http://www.tequipment.net/Velleman_PCSGU250.html -- also has a function generator
http://www.amazon.com/dp/B0007R8ZCG/?tag=tec06d-20
http://www.amazon.com/dp/B0036FZRU4/?tag=tec06d-20
All of which are under $400 (most <$250) and they're all new! You can get them for cheaper, though used, if you keep your eye on craigslist.

Also the quick start guide for the DS1M12 is 2 pages (maybe 100 words).

Using a USB oscope should take roughly the same amount of time as it would to make similar measurements with a multimeter. Using the information gleaned from the oscope, any laymen can determine if their PSU is pushing the limits of the ATX specification. It could be argued that the time it takes to post your multimeter measurements online, so that self-proclaimed experts can give you an incomplete answer, is considerably more than it would be if you'd used an inexpensive oscope.

Furthermore with so many TPU members buying computers every 2 years (or so) for around $1000 (give or take), it seems strongly advisable to buy a used and/or inexpensive oscope.

@Frick - "usable oscilloscopes", as opposed to what? Is an "unusuable" oscope a broken one?

 

[Frick]

Aye, usable oscillosopes you use for these things are pretty cheap considering what you get.

 

[westom]

I own several multimeters and I have actually tried to measure ripple with them.

No multimeter can measure ripple voltage directly. And need not do so. How excessive ripple voltage is detected is well beyond this discussion. And obvious when any layman measures those DC voltages even with a $17 Wal-Mart or K-mart meter.

You should know why a multimeter will not measure ripple voltage on AC settings (how it works). We don't try. Excessive ripple voltage is made obvious by how a multimeter and power supply works. Teaching how we designed and tested supplies is not relevant to layman. This is a layman discussion. How a layman can learn so much from one minute of labor.

Numbers from a meter posted here will immediately result in useful replies from those who better know this stuff. No numbers means those who know best can only post wild speculation - like so many others. Numbers from that meter identify operation of everything in a power 'system' - not just the supply. Massive information from only a minute of labor. From a tool that sells in most stores that also sell hammers.

Yes, an oscilloscope does make it easier for anyone to see failures such as ripple voltages. And to better learn how electricity works. But using a scope is not for people who are only expected to do what any 13 year old can do. A multimeter means anyone can have an immediate answer without wild speculation: without the 'it might be this' or 'could be that' answer.

 

[streetfighter 2]

westom, layman, I don't think this means what you think it means.

No multimeter can measure ripple voltage directly. And need not do so.

This is actually true.

How excessive ripple voltage is detected is well beyond this discussion.

No it isn't. It's been said like 20 times now. It's unbelievably simple:
Step 1) Connect cheap oscope to 12V
Step 2) Observe waveform and do a Vp-p measurement
Step 3) Check if waveform is exceeding 12V ±5% (as per ATX specification)
Step 4) . . .
Step 5) Profit.

That took like 6 seconds to explain . . .

Numbers from a meter posted here will immediately result in useful replies from those who better know this stuff.
. . .
A multimeter means anyone can have an immediate answer without wild speculation: without the 'it might be this' or 'could be that' answer.

Unfortunately you're wrong.

My measurements with various multimeters were only to amuse myself with how inaccurate of a reading I was getting. The most egregious voltage my multimeters measured was 11.6V, which is well within specifications. The oscope showed a completely different story, one which highlighted a dying PSU.

Simple facts:
1) The only way a multimeter can detect a malfunctioning PSU is when the average voltage drops or rises by an amount near or exceeding the ATX specification.
2) The majority of PSUs will cause instability/damage to the system proportionate to the PSU's load and ripple levels.
3) Ripple is nearly impossible to observe with a multimeter.

. . . using a scope is not for people who are only expected to do what any 13 year old can do.

That's what you think of the average TPU member? That's not very nice. Any 13 year-olds on this forum are enterprising, intelligent, and most of all resourceful. Better to give them the benefit of the doubt. Thirteen is the age of adulthood in many cultures.

 

[westom]

The most egregious voltage my multimeter measured was 11.6V, which is well within specifications.

When a layman reports 11.6 VDC from his meter, the reply says 12 volts is completely defective. That number must exceed 11.75 volts. Reasons why are found in how power supplies work and how a multimeter works combined with those spec numbers.

5 volts must exceed 4.87. 3.3 must exceed 3.23. If any voltage is that low, then a multimeter is reporting a defective output. 5% spec number is only one fact that defines that minimally acceptable number.

Your scope simply confirmed what a multimeter was reporting: a defective 12 volts.

Your attitude of having amateurs do more (ie use a scope) is admirable. Indeed, should be encouraged. But most computer repair people so fear anything as to also fear a tool sold in K-mart. Because (their attitude) K-mart sells some products that should only be used by a genius.

My experience is confirmed by what we see in the Silicon Valley. Americans are so technically naive and so fearful of learning that 60% of new Valley employees are immigrants from India and China. Finding people with technical knowledge (which means swallowing fears and other emotions) has become noticeably harder in the past 20 years.

Amazing are so many who think they might be electrocuted if probing voltages inside a computer. Technical knowledge even among some computer repairmen is that poor. Yes, that should worry everyone reading this.

Getting most to even use a multimeter - why do they fear so much when a cell phone is so much more complex? Technical ignorance and fear even among one's peers is widespread.

 

[streetfighter 2]

When the 13 year old reports 11.6 VDC from his meter, the answer is a completely defective 12 volts. That number must exceed 11.75 volts.

This is not accurate. My MAME machine reports 11.7V day in day out for years (and occasionally drops to around 11.6V). She purrs like a kitten.

It's not it "must exceed", it's it "damn well should exceed".

My current PSU is at 12.54V and I've yet to see this computer crash from anything I didn't accidentally do (like pushing the ram too hard).

5 volts must exceed 4.87. 3.3 must exceed 3.23. If any voltage is that low or lower, then a multimeter reports a defective output. 5% spec number is only one fact that defines minimally acceptable numbers.

rail|tolerance|range|ripple Vp-p
+5 VDC |±5% (±0.25 V) |+4.75 V to +5.25 V | 50 mV
+12 VDC | ±5% (±0.60 V)|+11.40 V to +12.60 V |120 mV
+3.3 VDC|±5% (±0.165 V)|+3.135 V to +3.465 V|50 mV

Many motherboards (and complementary hardware) are actually more generous than that and will give you some slack. On the other hand some are very sensitive and will crash quickly when the limits of the standard are probed. The multimeter is rarely the final verdict, and it's not the only tool needed for testing a power supply.

Your scope simply confirmed what a multimeter was reporting: a defective 12 volts.

Unfortunately you are wrong. It was reporting acceptable voltage, the ripple as seen by oscope was well out of parameter though. That PSU had a bad input filter which lead to large ripple. The PSU also had poor quality switching due to lower quality MOSFETs and controller, which resulted in undesirable voltage drop. It's sibling, with a working input filter cap was otherwise an acceptable power supply (currently used in my MAME machine) despite the voltage drop.

voltage droops and ripple are both ways to tell if somethings wrong with a PSU.

I never said voltage drops aren't a way to tell that something is wrong with a PSU, but it's not always (or even often) the final answer. As you said yourself some hardware is more/less sensitive to voltage and ripple conditions, which doesn't necessarily mean the PSU is defective.

I posted this earlier:
http://www.hardwaresecrets.com/article/Why-99-Percent-of-Power-Supply-Reviews-Are-Wrong/410/1
I find it to be more or less accurate.

 

[Mussels]

streetfighter 2: no he isnt, most PSU's try and remain within 3%.


just because you have systems that work fine outside those specs doesnt mean the specs have changed.


I've seen systems still booting and running at 11.2V, and i've seen others that crashed if they went below 11.9v


voltage droops and ripple are both ways to tell if somethings wrong with a PSU.

 

[Wile E]

streetfighter 2: no he isnt, most PSU's try and remain within 3%.


just because you have systems that work fine outside those specs doesnt mean the specs have changed.


I've seen systems still booting and running at 11.2V, and i've seen others that crashed if they went below 11.9v


voltage droops and ripple are both ways to tell if somethings wrong with a PSU.

Not unless you measures the psu before problems occured, and the voltages were different.

If you never took a meter to your psu befre you encountered problems, getting a 11.6v reading on the 12v does not show a sign of a bad psu. What if it read 11.6v prior to having issues, but you just never measured it?

The only way a multimeter proves anything is if your psu is reading different compared to when (if) you took a baseline, or it shows the psu running out of ATX specs (which is ±5% on the 12v rails or exactly what street posted).

 

[Mussels]

I posted this earlier:
http://www.hardwaresecrets.com/article/Why-99-Percent-of-Power-Supply-Reviews-Are-Wrong/410/1
I find it to be more or less accurate.

i missed that earlier, but its quite a good read.

I used to do PSU reviews, and quit for the same reasons they stated there - you need proper equipment to test it, and PSU companies didnt want you to. they sent guidelines on what they wanted in the reviews, and it really did consist of nothing more than plugging it into a PC and seeing if the voltages moved.


edit for wile E: i said in an earlier post that measuring them early on for a baseline was a good idea. You are correct in street posting the ±5% - i was agreeing on that point. a multimeter can let you know if its out of spec there, while an oscilliscope lets you know if its a shit PSU or not.

 

[bokou]

I've got a Xion 800W modular and I've been satisfied with it so far for about 2 years now. It's powered a 9800 GX2 and a 6970 and never given me trouble powering or staying on nearly 24/7 despite it being a 4-rail.

My next PSU will likely be a Corsair or some other reputable brand and will need to be 1000W and single-rail to withstand expandability. I doubt my Xion will run two 6970's in Xfire since it's a 4-rail and won't want to chance it. Just offering up my $.02 on this particular PSU.

 

[grimzz R]

what PSU should i get for this build?

Once You Know, You Newegg

 

[cdawall]

what PSU should i get for this build?

Once You Know, You Newegg

a good 500w would power it just fine

CORSAIR Builder Series CMPSU-500CX 500W ATX12V v2....

PC Power and Cooling Silencer Mk II PPCMK2S500 500...

PC Power & Cooling Silencer PPCS500D 500W ATX12V /...

OCZ ModXStream Pro OCZ500MXSP 500W ATX12V V2.2 / E...

 

[grimzz R]

i decided on the Corsair 600W.

 

[Mussels]

i decided on the Corsair 600W.

the CX is their budget (worst) line, but that doesnt mean it wont fit your needs perfectly. decent warranty on it too.

 

[bokou]

Just bought a Corsair 1200w and it's made by Flextronics. The OP could use to be updated a wee bit.

here's a review of the Corsair AX1200 - http://www.hardocp.com/article/2010/06/21/corsair_ax1200_1200w_power_supply_review/

 

[bokou]

bump - can we get Flextronics added as a supplier for Corsair?

Flextronics International Ltd. is an electronics manufacturing services provider, which makes it a bit atypical from the general consumer SMPS OEM, founded in 1969 and headquartered in Singapore. Flextronics provides a wide range of manufacturing, supply chain management and procurement, logistical support, and technical support services for industries ranging from automotive, to medical, to consumer electronic, and industrial component production. Some of their more familiar customers that they provide manufacturing services, or outright production for are familiar to many users including Lenovo, Lego Group, Microsoft, RIM, LG, SUN, HP, Kodak, Verizon, Amazon, Samsung, Siemens, Apple, and many more. Flextronics subsidiary FlexPower, was founded in 2005 and Flextronics claims that following the acquisition of Friwo and Coldwatt, plus internal growth and development that they are now the "4th largest power supply company in the world". However, its production of products for the consumer SMPS market is still rather limited with Corsair bringing the first product from Flextronics/FlexPower to the enthusiast we have seen to date.

 

[p_o_s_pc]

bump - can we get Flextronics added as a supplier for Corsair?

The poster of the OP isn't very active anymore an update may not happen. This thread still is useful for people as the later post give some updated info.

 

[bokou]

gotcha, good to know!